Repository navigation
feat(agents): migrate APEX to the Copilot SDK harness - #767
Merged
Merged
Conversation
- Archive Local entrypoints and adopt native Agent Host interactions - Add fail-closed worker readiness and SDK telemetry acceptance - Preserve project artifacts, human gates and specialist-owned models Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| argument-hint: Deploy the Bicep templates for a specific project | ||
| user-invocable: true | ||
| disable-model-invocation: true | ||
| disable-model-invocation: false |
| argument-hint: Deploy the Terraform configuration for a specific project | ||
| user-invocable: true | ||
| disable-model-invocation: true | ||
| disable-model-invocation: false |
| argument-hint: "Provide the path to the artifact to challenge (e.g. agent-output/my-project/04-implementation-plan.md)" | ||
| user-invocable: true | ||
| disable-model-invocation: true | ||
| disable-model-invocation: false |
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
Obsolete vendor-prompting assertions deterministically fail and must be updated for the SDK contract.
1 open finding
What changed in this PR
Migrates APEX from Local prompt adapters to the Copilot SDK harness, preserving specialist models and human approval gates.
Changes:
- Archives legacy adapters and introduces canonical entry skills.
- Adds verified worker readiness, native interactions, and telemetry contracts.
- Updates validators, registries, documentation, and regression tests.
Critical blocker: tools/tests/vendor-prompting/run.test.mjs:614 retains obsolete Local/Host assertions, while the preceding harness-boundary test also checks removed wording. These deterministically fail npm run test:vendor-prompting. Update both tests for SDK/readiness contracts, retaining legacy checks only for archived instructions.
Live VS Code acceptance and external documentation remain pending.
| File | Description |
|---|---|
tools/tests/vendor-prompting/run.test.mjs |
Redirects archive checks; obsolete assertions remain. |
tools/tests/scripts/test_validate_review_ceiling.py |
Tests unknown budget evidence. |
tools/tests/scripts/test_validate_question_batching.py |
Updates schema diagnostics. |
tools/tests/scripts/test_validate_orchestrator_handoff.py |
Tests SDK readiness and completion contracts. |
tools/tests/scripts/test_skill_ownership.mjs |
Checks adapter archival. |
tools/tests/scripts/test_skill_consolidation.mjs |
Updates manual skill identities. |
tools/tests/scripts/test_sdk_entrypoints.mjs |
Adds entrypoint and dispatch checks. |
tools/tests/scripts/test_review_presence.mjs |
Updates resume and approval checks. |
tools/tests/scripts/test_post_write_validation.mjs |
Checks SDK context boundaries. |
tools/tests/scripts/test_phase_a_runtime_guidance.mjs |
Uses native event fixtures. |
tools/tests/scripts/test_native_log_parser.py |
Tests scoped, read-only log analysis. |
tools/tests/scripts/test_lint_commands.mjs |
Tests retired entries and deleted paths. |
tools/tests/scripts/test_guidance_remediation.mjs |
Updates readiness assertions. |
tools/tests/scripts/test_execution_subagent_contract.mjs |
Tests canonical execution references. |
tools/tests/scripts/test_context_baseline.mjs |
Removes retired snapshot targets. |
tools/tests/scripts/test_agent_body_remediation.mjs |
Updates invocation and worker-boundary checks. |
tools/tests/lib/npm-script-graph.test.mjs |
Expects empty prompt roots. |
tools/tests/generate-explorer-graph.test.mjs |
Tests retired prompt exclusion. |
tools/tests/bats/tool-guardian.bats |
Adds SDK tool-input regressions. |
tools/tests/bats/subagent-validation.bats |
Adds SDK response regressions. |
tools/scripts/validate-skills.mjs |
Updates resume replacement guidance. |
tools/scripts/validate-orphaned-content.mjs |
Removes retired corpus input. |
tools/scripts/validate-no-hardcoded-counts.mjs |
Excludes archives. |
tools/scripts/validate-models.mjs |
Removes retired prompt roots. |
tools/scripts/validate-mcp-tool-refs.mjs |
Removes retired prompt scanning. |
tools/scripts/validate-instruction-checks.mjs |
Narrows active reference scanning. |
tools/scripts/validate-explorer-graph.mjs |
Removes retired source directories. |
tools/scripts/validate-agents.mjs |
Enforces SDK dispatch topology. |
tools/scripts/validate_review_ceiling.py |
Counts native executions and unknown evidence. |
tools/scripts/validate_question_batching.py |
Validates native form properties. |
tools/scripts/snapshot-agent-context.sh |
Snapshots active sources only. |
tools/scripts/safe-shell.mjs |
Updates scanning and interaction guidance. |
tools/scripts/lint-json.mjs |
Excludes archived JSON. |
tools/scripts/generate-explorer-graph.mjs |
Uses canonical prompt roots. |
tools/scripts/diff-context-baseline.sh |
Removes retired comparison categories. |
tools/scripts/check-repository-links.mjs |
Skips deleted cached paths. |
tools/scripts/_lib/paths.mjs |
Disables active prompt discovery. |
tools/schemas/model-catalog.schema.json |
Adds expected runtime model IDs. |
tools/schemas/agent-scorecard.schema.json |
Updates assessment reference documentation. |
tools/registry/count-manifest.json |
Removes prompt inventory definition. |
tools/registry/agent-registry.json |
Removes prompt registrations. |
README.md |
Documents SDK entrypoints and acceptance limits. |
package.json |
Adds worker-readiness command. |
eslint.config.mjs |
Excludes archives. |
CHANGELOG.md |
Records migration and compatibility changes. |
AGENTS.md |
Updates runtime and dispatch boundaries. |
.github/terminology-blocklist.json |
Excludes historical terminology. |
.github/skills/README.md |
Updates manual entry names. |
.github/skills/apex-workflow-start/SKILL.md |
Adds manual workflow entrypoint. |
.github/skills/apex-workflow-engine/SKILL.md |
Integrates readiness-based routing. |
.github/skills/apex-workflow-engine/references/workflow-entry.md |
Defines SDK ownership and resume behavior. |
.github/skills/apex-workflow-engine/references/subagent-integration.md |
Updates worker hierarchy guidance. |
.github/skills/apex-workflow-engine/references/review-imported-iac.md |
Removes adapter routing; updates input tool. |
.github/skills/apex-workflow-engine/references/project-wide-review.md |
Inventories skills instead of prompts. |
.github/skills/apex-workflow-engine/references/operational-safety.md |
Updates SDK operational boundaries. |
.github/skills/apex-workflow-engine/references/execution-subagent.md |
Updates execution authority wording. |
.github/skills/apex-iac-common/references/placeholder-scan-protocol.md |
Renames user-input tool. |
.github/skills/apex-iac-common/references/iac-planner-approval-gate.md |
Renames approval input tool. |
.github/skills/apex-iac-common/references/deploy-validation-checklist.md |
Updates placeholder input guidance. |
.github/skills/apex-iac-common/references/deploy-shared-workflow.md |
Updates deployment question tooling. |
.github/skills/apex-iac-common/references/codegen-shared-workflow.md |
Updates blocker and decision tooling. |
.github/skills/apex-iac-common/references/codegen-do-dont.md |
Updates question-tool references. |
.github/skills/apex-github-operations/SKILL.md |
Links renamed commit entrypoint. |
.github/skills/apex-github-operations/references/merge-sensei-free-pr.md |
Removes retired authoring references. |
.github/skills/apex-github-operations/references/git-commit.md |
Updates PR decision tooling. |
.github/skills/apex-git-commit/SKILL.md |
Adds manual commit entrypoint. |
.github/skills/apex-debug-log-export/SKILL.md |
Adds native-log export entrypoint. |
.github/skills/apex-context-management/SKILL.md |
Documents native analysis prerequisites. |
.github/skills/apex-context-management/references/context-audit.md |
Updates native log discovery guidance. |
.github/skills/apex-context-management/references/audit-setup.md |
Documents repository-scoped profiling setup. |
.github/skills/apex-azure-governance-discovery/references/reconciliation-disposition.md |
Renames finding-decision tool. |
.github/skills/apex-azure-governance-discovery/references/inline-resolution-gate.md |
Renames resolution input tool. |
.github/skills/apex-azure-governance-discovery/references/baseline-check.md |
Renames baseline-selection tool. |
.github/skills/apex-azure-defaults/references/vnet-planning.md |
Updates networking question-tool references. |
.github/skills/apex-azure-defaults/references/plan-design-decisions.md |
Updates planning question-tool references. |
.github/skills/apex-azure-defaults/references/cost-alerts-baseline.md |
Updates recipient-input tooling. |
.github/skills/apex-azure-artifacts/templates/01-requirements.template.md |
Links canonical requirements ownership. |
.github/skills/apex-agent-authoring/SKILL.md |
Exposes preserved assessment references. |
.github/skills/apex-agent-authoring/references/runtime-guardrails.md |
Adds verified execution prerequisites. |
.github/skills/apex-agent-authoring/references/plan-four-layer-agent-assessment.md |
Updates assessment references and terminology. |
.github/skills/apex-agent-authoring/references/assess-github-folder.md |
Removes retired adapter reference. |
.github/skills/apex-agent-authoring/references/assess-agents.md |
Updates rubric links and runtime terminology. |
.github/skills/apex-agent-authoring/references/agent-file-structure.md |
Documents SDK metadata boundaries. |
.github/model-catalog.json |
Adds expected SDK model identities. |
.github/instructions/sku-manifest.instructions.md |
Renames preference-input tool. |
.github/instructions/references/precedence-matrix.md |
Documents archived prompt instructions. |
.github/instructions/no-interactive-shell.instructions.md |
Updates scope and input-tool guidance. |
.github/instructions/markdown.instructions.md |
Removes retired prompt scope. |
.github/instructions/context-optimization.instructions.md |
Clarifies archived routing metadata. |
.github/instructions/agent-skills.instructions.md |
Defines SDK skill boundaries. |
.github/instructions/agent-operating-frame.instructions.md |
Updates delegation and checkpoint contracts. |
.github/instructions/agent-authoring.instructions.md |
Defines verified dispatch and native forms. |
.github/hooks/subagent-validation/subagent-validation.sh |
Adapts advisory validation to SDK responses. |
.github/copilot-instructions.md |
Defines SDK orchestration boundaries. |
.github/CODEOWNERS |
Moves ownership coverage to skills. |
.github/agents/09-diagnose.agent.md |
Adds delegated diagnostic entry contract. |
.github/agents/08-as-built.agent.md |
Adds worker entry and completion contracts. |
.github/agents/04-design.agent.md |
Adds worker entry and completion contracts. |
.github/agents/_subagents/terraform-validate-subagent.agent.md |
Requires verified execution handshake. |
.github/agents/_subagents/terraform-plan-subagent.agent.md |
Requires verified execution handshake. |
.github/agents/_subagents/policy-precheck-subagent.agent.md |
Requires verified execution handshake. |
.github/agents/_subagents/cost-estimate-subagent.agent.md |
Requires verified execution handshake. |
.github/agents/_subagents/challenger-review-subagent.agent.md |
Requires verified execution handshake. |
.github/agents/_subagents/bicep-whatif-subagent.agent.md |
Requires verified execution handshake. |
.github/agents/_subagents/bicep-validate-subagent.agent.md |
Requires verified execution handshake. |
.archive/pre-cpharness/tests/test_host_adapters.mjs |
Retains historical adapter tests. |
.archive/pre-cpharness/skills/apex-host-workflow-start/SKILL.md |
Archives Host workflow entrypoint. |
.archive/pre-cpharness/skills/apex-host-git-commit/SKILL.md |
Archives Host commit entrypoint. |
.archive/pre-cpharness/skills/apex-host-debug-log-export/SKILL.md |
Archives Host export entrypoint. |
.archive/pre-cpharness/README.md |
Explains archive boundaries. |
.archive/pre-cpharness/prompt.instructions.md |
Preserves historical prompt guidance. |
.archive/pre-cpharness/github-prompts/apex-resume-workflow.prompt.md |
Archives resume adapter. |
.archive/pre-cpharness/github-prompts/apex-git-commit.prompt.md |
Archives commit adapter. |
.archive/pre-cpharness/github-prompts/apex-debug-log-export.prompt.md |
Archives export adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/10-challenger-review.prompt.md |
Archives Challenger adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/09-diagnose-resource.prompt.md |
Archives diagnostic adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/08-as-built.prompt.md |
Archives as-built adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/07t-terraform-deploy.prompt.md |
Archives Terraform deployment adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/07b-bicep-deploy.prompt.md |
Archives Bicep deployment adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/06t-terraform-codegen.prompt.md |
Archives Terraform code-generation adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/06b-bicep-codegen.prompt.md |
Archives Bicep code-generation adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/05-iac-planner.prompt.md |
Archives planning adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/04g-governance.prompt.md |
Archives governance adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/04-design.prompt.md |
Archives design adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/03-architect.prompt.md |
Archives architecture adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/02-requirements.prompt.md |
Archives requirements adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/01-new-project.prompt.md |
Archives project intake adapter. |
.archive/pre-cpharness/apex-prompts/workflow-prompts/00-resume-workflow.prompt.md |
Archives attachable resume adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/review-imported-iac.prompt.md |
Archives imported-IaC adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/project-wide-review.prompt.md |
Archives project-review adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/plan-fourLayerAgentAssessment.prompt.md |
Preserves assessment planning history. |
.archive/pre-cpharness/apex-prompts/utility-prompts/merge-sensei-free-pr.prompt.md |
Archives merge adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/execution-subagent.prompt.md |
Archives execution adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/execution-subagent-gpt.prompt.md |
Archives GPT execution adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/eventdesk-orchestrator.prompt.md |
Preserves historical project intake. |
.archive/pre-cpharness/apex-prompts/utility-prompts/context-audit.prompt.md |
Archives context-audit adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/assess-github-folder.prompt.md |
Archives folder-assessment adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/assess-agents.prompt.md |
Archives agent-assessment adapter. |
.archive/pre-cpharness/apex-prompts/utility-prompts/as-built-from-azure.prompt.md |
Archives existing-deployment adapter. |
🧠 Review effort: Balanced
💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.
Comment on lines
614
to
615
| ".archive/pre-cpharness/prompt.instructions.md", | ||
| ".github/instructions/agent-skills.instructions.md", |
This was referenced Oct 17, 2026
Recover the original plan and add delegable delivery artifacts. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Resolve the Prettier failure reported by CI for issue #769. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This was referenced Oct 17, 2026
Apply formatter-owned table alignment without changing audit content. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Require linked override-free dispatch evidence for current SDK labels. Retain fail-closed readiness checks and add regression coverage. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Make the nonce-only ACK contract explicit in delegated main agent bodies. Retain tool-free stops and execution gates; cover the contract with regression tests. Refs #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Remove active model and effort pins and update coupled policy validators. Separate initial routing checks from complete same-worker readiness. Keep unsupported Auto telemetry blocked pending a fresh runtime capture. Refs #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
| argument-hint: Deploy the Terraform configuration for a specific project | ||
| user-invocable: true | ||
| disable-model-invocation: true | ||
| disable-model-invocation: false |
| argument-hint: Deploy the Bicep templates for a specific project | ||
| user-invocable: true | ||
| disable-model-invocation: true | ||
| disable-model-invocation: false |
Accept observed unpinned session inheritance only when worker and parent models agree. Align handshake parsing and prompts with captured SDK output; keep operator Auto unverified. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Verify named-worker identity, linked tool-free turns and the same-worker ACK. Treat models, effort, picker selection and model-source telemetry as diagnostics only. Update runtime guidance, smoke acceptance, provenance metadata and regression tests. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Owner
Author
Automatic review and technical handoffsPublished
Validation
Runtime acceptance remains separateNo live rubber-duck project review was invoked. The earlier Step 1 smoke still This change does not close #773 or certify full live smoke acceptance. |
9 of 10 tasks
Owner
Author
|
Final verification: pushed follow-up |
Accept omitted task mode only with attributed synchronous runtime proof. Validate internal tool turns within one worker interaction and preserve strict JSON output. Align telemetry and clarify reviewer input, output and request identity rules. Refs #767, #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Owner
Author
Live one-shot recorder repairPublished Confirmed runtime defects and fixes:
Validation: 870 integrated Node tests passed with 2 existing skips; 302 focused Read-only replay of the original native reviewer now passes the initial mode No new reviewer execution, project-state mutation, approval, budget reset, |
Owner
Author
|
Published budget-parser recovery in 0f94724. The original native log now parses unchanged: six attributed executions, one APEX_REVIEW; strict budget reports Step 1: 1, within ceiling 2. Parent WORKER_EXECUTE and child APEX_REVIEW IDs are namespaced by protocol, not collapsed or ignored. Same-protocol duplicate IDs still fail, and actual repeated dispatches still exceed the ceiling when appropriate. Regression validation: 153 focused Python tests passed; 810 Node tool-contract tests passed with 2 existing skips; Ruff, Markdown, skills, docs freshness and normal hooks passed. The latest pilot stopped before a new dispatch, so no new review attempt was consumed. Original malformed findings remain unaccepted; Gate 1 is closed. |
Owner
Author
|
Second actual one-shot attempt consumed the final default slot; Gate 1 remains closed. It returned fenced JSON with string traces_to values and invented finding IDs, and incurred an out-of-bounds reference read. No findings were repaired or accepted. The user explicitly selected Override ceiling for exactly one additional review after handoff hardening. Published 2baa5f2: typed output guidance and reviewer-owned in-memory ID/count computation plus schema preflight. The parent/recorder still cannot repair returned findings. Executable documentation tests and 812 Node tool-contract tests passed (2 existing skips); authoring/Markdown/skills/safe-shell/freshness checks and normal hooks passed. No extra live dispatch has been made by engineering. |
Comment on lines
7
to
+9
| user-invocable: true | ||
| disable-model-invocation: true | ||
| disable-model-invocation: false | ||
| agents: ["bicep-whatif-subagent", "bicep-validate-subagent", "policy-precheck-subagent"] | ||
| tools: | ||
| [vscode/askQuestions, execute, read, agent, edit, search, 'azure-mcp/*', todo] | ||
| [execute, read, agent, edit, search, 'azure-mcp/*', todo] |
Comment on lines
+6
to
+9
| disable-model-invocation: false | ||
| agents: ["terraform-plan-subagent", "terraform-validate-subagent", "policy-precheck-subagent"] | ||
| tools: | ||
| [vscode/askQuestions, execute, read, agent, edit, search, 'azure-mcp/*', todo] | ||
| [execute, read, agent, edit, search, 'azure-mcp/*', todo] |
- Enable direct WORKER_EXECUTE dispatch (--phase execute) - Make record-review.mjs resilient to fences, traces_to, IDs, and log flush - Fix profile_debug_log.py event ordering and per-worker first-input metrics - Complete PRD, backlog, smoke acceptance, and telemetry verification Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Owner
Author
Copilot Harness Migration — Autopilot Overhaul & Release Readiness (
|
jonathan-vella
added a commit
that referenced
this pull request
Oct 17, 2026
Restores main to pre-#767 tree while keeping work on feat/copilot-harness-migration. Co-authored-by: Jonathan Vella <jonathan-vella@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
jonathan-vella
added a commit
that referenced
this pull request
Oct 20, 2026
…on (#779) * feat(agents): migrate APEX to the Copilot SDK harness - Archive Local entrypoints and adopt native Agent Host interactions - Add fail-closed worker readiness and SDK telemetry acceptance - Preserve project artifacts, human gates and specialist-owned models Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: Preserve harness migration plan and delivery backlog Recover the original plan and add delegable delivery artifacts. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Format SDK smoke acceptance document Resolve the Prettier failure reported by CI for issue #769. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Align CI Python and repair archived audit link Provide the supported interpreter before Node tool contracts. Add regression coverage for #776 and repair the historical source link for #777. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Align historical audit table after link repair Apply formatter-owned table alignment without changing audit content. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Verify current SDK definition routing attribution Require linked override-free dispatch evidence for current SDK labels. Retain fail-closed readiness checks and add regression coverage. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Echo handshake identity in readiness acknowledgments Make the nonce-only ACK contract explicit in delegated main agent bodies. Retain tool-free stops and execution gates; cover the contract with regression tests. Refs #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * feat(agents): Use operator-selected Auto without model pins Remove active model and effort pins and update coupled policy validators. Separate initial routing checks from complete same-worker readiness. Keep unsupported Auto telemetry blocked pending a fresh runtime capture. Refs #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Verify inherited readiness telemetry Accept observed unpinned session inheritance only when worker and parent models agree. Align handshake parsing and prompts with captured SDK output; keep operator Auto unverified. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Make worker readiness model-agnostic Verify named-worker identity, linked tool-free turns and the same-worker ACK. Treat models, effort, picker selection and model-source telemetry as diagnostics only. Update runtime guidance, smoke acceptance, provenance metadata and regression tests. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * feat(agents): Automate one-shot rubber-duck reviews Keep technical returns with the Orchestrator and preserve approval and review gates. Record attributed independent results without reviewer readiness turns. Refs #767, #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * test(agents): Align frozen-input review metadata assertion Keep frozen Step 4 coverage tied to captured supporting-input metadata. Refs #767, #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Record native synchronous multi-turn reviews Accept omitted task mode only with attributed synchronous runtime proof. Validate internal tool turns within one worker interaction and preserve strict JSON output. Align telemetry and clarify reviewer input, output and request identity rules. Refs #767, #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(scripts): Scope execution identities by protocol Keep historical parent execution and child review requests independently counted. Preserve same-protocol duplicate rejection and strict review ceilings. Refs #767, #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Make reviewer draft schema explicit Require array traces and reviewer-owned in-memory ID and count generation. Test the documented preflight without changing returned findings or recorder guards. Refs #767, #773 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * feat(agents): Streamline Copilot SDK harness for Autopilot execution - Enable direct WORKER_EXECUTE dispatch (--phase execute) - Make record-review.mjs resilient to fences, traces_to, IDs, and log flush - Fix profile_debug_log.py event ordering and per-worker first-input metrics - Complete PRD, backlog, smoke acceptance, and telemetry verification Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Preserve truthful migration acceptance and review provenance Resolve reviewer IDs from an explicit session and exact request ID. Reject post-dispatch metadata repair and unverified tool outcomes. Verify direct worker receipts without claiming continuation was tested. Restore strict telemetry checks and reopen unsupported completion claims. Validate source separately from incomplete workspace project artifacts. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: Format migration recovery evidence table Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: Record verified owner-initiated test cleanup Keep source-validation results separate from incomplete E2E acceptance. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: Replan VS Code Copilot-only migration Revise the PRD, roadmap, issue packets and Autopilot execution contract. Preserve historical evidence and stop at owner plan approval. Refs #779, #780, #781, #782, #783, #784, #785, #786, #787, #788 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: Track additional migration review blockers Include diagnostic approval-channel coverage without implementing production fixes. Refs #779, #781, #782 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Retire runtime hooks and reconcile Copilot contracts Honor the owner-deleted hooks, remove stale callers and retain Git/security checks. Fix direct dispatch and manual review guidance; record genuine execution evidence. Full source-only validation passes; protected project and runtime/E2E gaps remain. Refs #779, #775, #781, #789 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Preserve manual entry for consequential agent roles Disable implicit invocation for deploy, diagnose and optional review entries. Keep credential isolation and actual approval enforcement explicitly unverified. Refs #779, #781, #782, #784 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * refactor(agents): Migrate all bodies to native Copilot execution Preserve direct questions, bounded delegated inputs and domain output contracts. Remove UI handoff and worker handshake dependencies; align guidance and tests. Full source-only validation passes; document actual remaining runtime probe gaps. Refs #779, #773, #782, #784, #786 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(ci): Accept native agents without legacy UI handoff references Retain invalid-target and empty-inventory failures with negative regression tests. Refs #779, #786 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: Record owner-observed native agent retests Distinguish seven passing question/input probes from untested Azure operations. Preserve prior observations and qualify user-supplied evidence provenance. Refs #779, #773, #782 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: Record direct question probes for every main agent Preserve user-reported provenance and separate synthetic interaction from approval. Refs #779, #773, #782 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Continue selected Orchestrator into requirements discovery Distinguish initialization-only requests and avoid a redundant role-selection stop. Preserve existing project state and unconfirmed template defaults. Refs #779, #772, #782 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Restore guided Requirements discovery Preserve complete answer ledgers, conditional menus and automatic native review relay. Keep genuine review evidence and deployment approval gates unchanged. Refs #772, #782, #784, #786 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Remove shell-based reviewer finalization Return substantive findings directly and reuse deterministic recorder normalization. Preserve failed-command rejection, review freshness and existing attempt authority. Clarify Requirements review scope without changing historical project findings. Refs #772, #784, #786 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(instructions)!: Remediate audit and enforce Copilot contracts Align governance, artifact, shell and documentation guidance with actual contracts. Retire Local-only discovery settings and add pinned PowerShell analysis, hook integration and regression coverage. BREAKING CHANGE: remove fix:artifact-h2; use fix:artifacts with explicit --apply. Local-only customization location settings are no longer accepted. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Preserve negated scenario choices in discovery Handle negated Greenfield and coordinated migration or modernization terms. Add seven regressions and record the zero-credit deterministic validation pass. Preserve project artifacts, approvals and review-attempt history. Refs #772, #782, #786 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(agents): Preserve native question form wrappers Reject full JSON Schema roots and bound correction without re-asking captured answers. Record actual CLI evidence and preserve the prose-review recording blocker. Keep recorder integrity, review budgets, approvals and existing project artifacts unchanged. Refs #772, #773, #784, #786 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * refactor(agents): Share one structured review backend Use the existing read-only typed leaf for automatic and optional manual review. Record actual native output with definition and procedure hashes; preserve historical readers. Align callers, skills, instructions, gate diagnostics and invocation accounting. Fresh CLI recording and strict validation pass; fresh VS Code context acceptance remains blocked. Refs #772, #773, #784, #786 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(scripts): Report structured reviewer execution failures Keep execution-failure JSON out of findings and surface its exact stage and error. Preserve historical extraction and leave failed receipts unpublished. Refs #784 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Resolve workflow contract drift and validator blind spots across repo - Align decision-keys.md complexity enum and register workflow toggles - Sync top-level iac_tool/region in apex-recall decide/transition/show - Validate multi-pass and simple-tier deep reviews in complete-step - Document subnet-plan and lesson-log enum constraints in instructions - Fix false negatives in decision-keys, lens-references, and context-budget - Fix in-progress step handling in validate-challenger-presence - Wire standalone contract validators and tests into validate:_node suites - Rebuild architecture-explorer-graph.json and eliminate skill lint warnings Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: Apply cp-harness-01 lessons learned to subagents, skills, and trace validator - Forbid speculative view calls on unverified paths in reviewer (LL-001) - Document Storage restorePolicyDays < deleteRetentionPolicyDays rule (LL-002) - Accept not-applicable/auto-remediated rows in governance trace check (LL-003) - Emphasize kebab-case pricing_get args in cost-estimate-subagent (LL-004) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * chore: apply dependabot updates, align governance collector, and archive migration plans - Incorporate PR #756 (.devcontainer/devcontainer-lock.json node feature 2.1.1). - Incorporate PR #765 devDependencies (eslint 10.12.0, eslint-plugin-n 18.4.1, js-yaml 5.4.3, jsdom 30.1.2, lefthook 2.2.1, prettier 3.9.10, whatwg-url 17.2.0) compatible with the 7-day Microsoft CFS packagefeedproxy.microsoft.io quarantine window. - Incorporate PR #766 (ruff 0.17.0 in requirements.txt) and configure a 7-day Dependabot cooldown for npm and pip ecosystems in .github/dependabot.yml. - Align tools/scripts/collect-governance-baseline.ps1 with discover.py parameter resolution, conditional effect evaluation, and PARTIAL discovery status reporting (#763). - Update README.md for challenger-review-subagent one-shot review recording and archive completed migration planning packets into .archive/post-cpharness-cleanup-20261019.tar.gz. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(scripts): Support downstream apex-accelerator repositories in validation suite - Fall back to governance-policy-baseline.fixture.json in validate-governance-baseline.mjs when the private compressed baseline has not yet been collected in a template clone. - Ignore upstream-only .github/workflows/* node paths in validate-explorer-graph.mjs when running in a downstream consumer repository without .github/workflows/ci.yml. - Guard upstream-only git history, workflow, and mermaid diagram assertions in audit-retirement-candidates.test.mjs, compare-devcontainer-verdicts.test.mjs, test_instruction_audit.mjs, and test_lint_commands.mjs so npm run validate:all passes cleanly in synced apex-accelerator repositories. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Jonathan Vella <jonathan-vella@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
independent reviews and artifact boundaries.
Auto and HydraFusion. Models, effort, picker identifiers and model-source labels are diagnostics only.
same-worker continuation. Leave per-call routing overrides unset.
Readiness contract change
HANDSHAKE_VERIFIED, notMODEL_VERIFIED; it permits onlythe nonce-only continuation, not project execution.
READY_VERIFIEDandcontinuation_verified: trueafter the same worker's exact identity-bearing ACK and completed second turn.
model_observations; they cannot block readiness.closed capture, scope limits, recovery and review evidence remain mandatory.
runtime-managed; prior reviews are not restamped or implicitly refreshed.Validation of 33c9d88
HANDSHAKE_VERIFIED.This alone does not establish complete readiness or authorize execution.
Remaining acceptance gates
native input, independent review and human-stop acceptance matrix.
No Azure deployment or project-state mutation is included. Passing source checks or readiness
does not establish complete runtime workflow acceptance.