Migrate ai-agents from PowerShell to Python as primary scripting
language over a 12-24 month phased migration period.

## Decision Summary

- Python 3.10+ established as project language standard
- ADR-005 superseded for new development
- Phased approach: Foundation -> New Development -> Migration
- Python already prerequisite via skill-installer (PR #962)

## Rationale

- 70-second PowerShell tool startup times per invocation
- No CodeQL support for PowerShell (deterministic security unavailable)
- AI/ML ecosystem (Anthropic SDK, MCP) is Python-native
- skill-installer already requires Python 3.10+ and UV

## 6-Agent ADR Review Debate

| Agent | Verdict |
|-------|---------|
| Analyst | CONCERNS |
| Architect | CONCERNS |
| Critic | CONCERNS |
| Independent-Thinker | CONCERNS |
| Security | CONCERNS |
| High-Level-Advisor | ACCEPT |

Result: Disagree-and-Commit (5 CONCERNS + 1 ACCEPT)
Tie-breaker: High-Level-Advisor

## P0 Issues Resolved

- Stack Overflow claim corrected (Python growth, not #1)
- Path Dependence language fixed ("Python-first with phased migration")

## P1 Issues Deferred to Phase 1 Implementation

- pyproject.toml creation
- pytest infrastructure setup
- PROJECT-CONSTRAINTS.md update
- Supply chain controls (uv.lock, Dependabot, pip-audit)

BREAKING CHANGE: ADR-005 PowerShell-only standard superseded.
New scripts SHOULD be Python. Existing scripts migrate incrementally.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Migrate ai-agents from PowerShell to Python as primary scripting
language over a 12-24 month phased migration period.

## Decision Summary

- Python 3.10+ established as project language standard
- ADR-005 superseded for new development
- Phased approach: Foundation -> New Development -> Migration
- Python already prerequisite via skill-installer (PR #962)

## Rationale

- 70-second PowerShell tool startup times per invocation
- No CodeQL support for PowerShell (deterministic security unavailable)
- AI/ML ecosystem (Anthropic SDK, MCP) is Python-native
- skill-installer already requires Python 3.10+ and UV

## 6-Agent ADR Review Debate

| Agent | Verdict |
|-------|---------|
| Analyst | CONCERNS |
| Architect | CONCERNS |
| Critic | CONCERNS |
| Independent-Thinker | CONCERNS |
| Security | CONCERNS |
| High-Level-Advisor | ACCEPT |

Result: Disagree-and-Commit (5 CONCERNS + 1 ACCEPT)
Tie-breaker: High-Level-Advisor

## P0 Issues Resolved

- Stack Overflow claim corrected (Python growth, not #1)
- Path Dependence language fixed ("Python-first with phased migration")

## P1 Issues Deferred to Phase 1 Implementation

- pyproject.toml creation
- pytest infrastructure setup
- PROJECT-CONSTRAINTS.md update
- Supply chain controls (uv.lock, Dependabot, pip-audit)

BREAKING CHANGE: ADR-005 PowerShell-only standard superseded.
New scripts SHOULD be Python. Existing scripts migrate incrementally.

Co-authored-by: Test <test@test.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>

* docs!: add ADR-042 Python migration strategy (supersedes ADR-005)

Migrate ai-agents from PowerShell to Python as primary scripting
language over a 12-24 month phased migration period.

## Decision Summary

- Python 3.10+ established as project language standard
- ADR-005 superseded for new development
- Phased approach: Foundation -> New Development -> Migration
- Python already prerequisite via skill-installer (PR #962)

## Rationale

- 70-second PowerShell tool startup times per invocation
- No CodeQL support for PowerShell (deterministic security unavailable)
- AI/ML ecosystem (Anthropic SDK, MCP) is Python-native
- skill-installer already requires Python 3.10+ and UV

## 6-Agent ADR Review Debate

| Agent | Verdict |
|-------|---------|
| Analyst | CONCERNS |
| Architect | CONCERNS |
| Critic | CONCERNS |
| Independent-Thinker | CONCERNS |
| Security | CONCERNS |
| High-Level-Advisor | ACCEPT |

Result: Disagree-and-Commit (5 CONCERNS + 1 ACCEPT)
Tie-breaker: High-Level-Advisor

## P0 Issues Resolved

- Stack Overflow claim corrected (Python growth, not #1)
- Path Dependence language fixed ("Python-first with phased migration")

## P1 Issues Deferred to Phase 1 Implementation

- pyproject.toml creation
- pytest infrastructure setup
- PROJECT-CONSTRAINTS.md update
- Supply chain controls (uv.lock, Dependabot, pip-audit)

BREAKING CHANGE: ADR-005 PowerShell-only standard superseded.
New scripts SHOULD be Python. Existing scripts migrate incrementally.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* docs(planning): add ADR-042 Python migration implementation plan

Self-contained 618-line plan synthesizing inputs from:
- traycerai[bot]: Phase structure validation
- coderabbitai[bot]: 9 actionable suggestions
- github-actions[bot]: Detailed PRD with success metrics

Covers:
- Phase 1: Foundation (pyproject.toml, pytest, security controls)
- Phase 2: New Development Guidelines
- Phase 3: Migration (priority order, deprecation timeline)

Complete code templates included for immediate execution.

Relates-to: #965

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* docs(planning): add verification sections for autonomous execution

Enhance ADR-042 implementation plan for amnesiac agent execution:
- Add Quick Verification section with pre-flight checks
- Add Session Protocol section with JSON template
- Add Local File References table (all verified 2026-01-18)
- Add repository field to header metadata

Plan now 712 lines, fully self-contained for context-free execution.

Relates-to: #965

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* feat(python)!: implement Phase 1 Python infrastructure (ADR-042)

BREAKING CHANGE: Language policy changes from PowerShell-only to Python-first

Phase 1 establishes Python infrastructure for new development:

Infrastructure:
- pyproject.toml: Project metadata, dependencies, tool configs (ruff, mypy, pytest)
- uv.lock: Hash-pinned dependencies for supply chain security (16 packages)
- tests/conftest.py: Shared pytest fixtures (project_root, temp_test_dir)
- .github/workflows/pytest.yml: CI workflow with paths-filter, coverage, pip-audit, bandit

Policy Updates:
- PROJECT-CONSTRAINTS.md: SHOULD prefer Python for new scripts (ADR-042)
- CRITICAL-CONTEXT.md: Python-first (.py preferred)
- .githooks/pre-commit: Non-blocking Python linting with ruff
- .github/dependabot.yml: pip ecosystem for dependency updates

Housekeeping:
- .gitignore: Python patterns (__pycache__, .venv, .egg-info, etc.)
- .markdownlint-cli2.yaml: Exclude .venv from linting

Verification: uv pip install -e ".[dev]" succeeds, pytest discovers 77 tests

Refs: #965, ADR-042

Co-Authored-By: Claude <noreply@anthropic.com>

* docs: update documentation for Python-first development (ADR-042)

Update CONTRIBUTING.md and AGENTS.md to reflect the Python migration:

- Change "Always Do" from PowerShell-only to Python-first for new scripts
- Update "Never Do" to prohibit bash only (Python now allowed)
- Add Python 3.12.x and UV to Tech Stack table
- Add pytest testing section with automated quality gates emphasis
- Update Development Tools commands to include Python testing
- Emphasize shift-left automation: pre-commit hooks and CI handle quality
- Note Python 3.12.x requirement due to Ubuntu 25 incompatibility

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* docs(session): update session log with documentation changes

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* feat(python): implement Phase 2 parallel infrastructure (ADR-042)

Add documentation and security utilities for Python development:

- Create CI/CD migration patterns guide for GitHub Actions integration
- Create Python security checklist covering CWE-22, CWE-78, CWE-798
- Create path validation utility with 42 tests for CWE-22 protection
- Create PowerShell-to-Python developer migration guide

Part of epic #965.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* feat(python): add pilot migration of Check-SkillExists to Python (ADR-042 Phase 3)

Migrates Check-SkillExists.ps1 to Python as the pilot script for ADR-042 Phase 3.
This demonstrates the migration patterns established in Phase 2.

Changes:
- scripts/check_skill_exists.py: Python port with argparse CLI, type hints,
  ADR-035 exit codes, and path_validation utility usage
- tests/test_check_skill_exists.py: 31 pytest tests with 88% coverage

The Python version provides:
- --list-available: Lists all skills by operation type
- --operation/--action: Checks if a skill exists using substring matching
- --project-root: Optional custom project root for testing

Both PowerShell and Python versions will run in parallel per migration plan.

Refs: #965

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* feat(python): add gradual rollout migrations (ADR-042 Phase 4)

Migrate two additional scripts from PowerShell to Python following
the pilot pattern established in Phase 3:

- Detect-SkillViolation.ps1 -> detect_skill_violation.py
  - 89% test coverage (35 tests)
  - Uses dataclass for Violation type
  - Integrates path_validation utility
  - Non-blocking warning for skill violations

- Validate-SessionJson.ps1 -> validate_session_json.py
  - 91% test coverage (39 tests)
  - Uses ValidationResult dataclass
  - Case-insensitive JSON key lookup
  - Pre-commit mode for compact output

Also fixes uv.lock format (was incorrectly in pip-tools format,
now in native uv format).

See: ADR-042 Python Migration Strategy, Issue #965

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(python): fix regex bug and dead code in detect_skill_violation

- Fixed regex pattern gh\\s\+ to gh\s+ in extract_capability_gaps
- Replaced duplicated capability extraction logic in report_violations
  with call to extract_capability_gaps function (DRY)
- All 34 tests pass

Issues identified by pr-review-toolkit parallel review agents.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(security): address gemini-code-assist security findings

**Security Fixes**:

1. **Path Validation (CWE-22)** - scripts/validate_session_json.py:
   - Added `validate_safe_path` import from scripts.utils.path_validation
   - Validate user-provided session_path before file operations
   - Prevents path traversal attacks (../, symlinks, etc.)

2. **Python Version Alignment** - pyproject.toml:
   - Updated ruff target-version: py310 → py312
   - Updated mypy python_version: 3.10 → 3.12
   - Aligns linting/type checking with project standard (3.12.x)

**Gemini Review Comments Addressed**:
- Comment 2702879539: Added path validation imports ✓
- Comment 2702879541: Added CWE-22 protection with validate_safe_path ✓
- Comment 2702879542: Updated ruff to target py312 ✓
- Comment 2702879543: Updated mypy to python 3.12 ✓

**Testing**:
- Verified imports work correctly
- Path validation prevents traversal attacks
- Session protocol validation: PASS

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* refactor: slim instructions files by removing redundant agent registry (#961)

Remove agent catalog tables and routing heuristics from instruction
file templates. This content is already available in YAML frontmatter
of each agent file, which platforms parse directly.

- Claude: 129 → 45 lines (65% reduction)
- Copilot CLI: 126 → 53 lines (58% reduction)
- VSCode: 116 → 45 lines (61% reduction)

Estimated savings: ~2,000 tokens per session per platform.

Signed-off-by: Richard Murillo <6811113+rjmurillo@users.noreply.github.com>
Co-authored-by: Richard Murillo <richard.murillo@example.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Richard Murillo <6811113+rjmurillo@users.noreply.github.com>

* docs(analysis): Factory-AI/droid-action security constraint blocker (#960)

* docs(analysis): document Factory-AI/droid-action security constraint blocker

Root cause analysis of Droid Auto Review workflow failure. The Factory-AI/droid-action
internally uses actions/upload-artifact@v4 (non-SHA-pinned), which violates repository
security constraints requiring all actions to be pinned to full-length commit SHAs.

Key findings:
- Latest droid-action version (e3f8be9f, 2026-01-12) still contains non-pinned references
- Repository security rules apply recursively to all nested action dependencies
- No workaround available without modifying third-party action or relaxing security constraints

Impact: BLOCKING - droid-review.yml and droid.yml workflows fail at setup phase

Recommendations:
- File issue with Factory-AI requesting SHA-pinned action references
- Evaluate alternative PR review automation tools
- Document as known limitation in operational runbook

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* docs(session): update session-9 with PR comment responses

- Added workLog entries for PR #960 comment activities
- Documented upstream issue research (no issue exists)
- Added PR comments as deliverables
- Added learning pattern about upstream issue verification

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* docs: record upstream issue Factory-AI/droid-action#20

- Updated memory with upstream issue link and status
- Marked "file upstream issue" as DONE in recommendations
- Updated session log with issue filing activity
- Added next step to monitor for maintainer response

Upstream: Factory-AI/droid-action#20

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: rjmurillo-bot <rjmurillo-bot@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Co-authored-by: Richard Murillo <richard.murillo@example.com>

* fix(ci): disable Droid workflows due to unpinned action (#957)

* chore: recover 650 orphaned session logs and memory files (#964)

* chore: recover 650 orphaned session logs and memory files

Extract artifacts from 52 feature branches that were left behind when
PRs auto-merged before session logs were pushed.

Recovery summary:
- Session logs: 378 files recovered
- Memory files: 272 files recovered
- Total: 650 files, 82,632 lines of content

Analysis found 61,497 file references across branches but only 1,728
unique files (average file in 35+ branches). Of these, 1,080 already
existed in main. The 648 truly orphaned files are now consolidated.

Used consolidated PR approach instead of 52 individual PRs to avoid
massive merge conflicts from overlapping content.

Note: 150 memory files use legacy 'skill-' prefix naming that predates
ADR-017. These are historical artifacts being preserved as-is.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* docs: update session log with PR #964 details

Add PR information and audit trail for validation skip.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

---------

Co-authored-by: rjmurillo-bot <noreply@github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

* feat: implement investigation-only session validator (ADR-034 Phase 1) (#931)

* Initial plan

* Add comprehensive test suite for investigation-only validation

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* fix: convert functional tests to pattern-based tests to avoid git state dependency

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* fix(validation): Allow .agents/memory/ in investigation-only sessions (#926)

* Initial plan

* feat: Add .agents/memory/ to investigation allowlist

Add .agents/memory/ pattern to investigation-only allowlist in Test-InvestigationEligibility.ps1 scripts and update tests. This allows memory infrastructure files like causal-graph.json to be committed in investigation sessions per ADR-034 memory-first principle.

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* feat: Add verification-based session-start gates for Codex effectiveness (#924)

* Initial plan

* docs: add Codex effectiveness backlog and context optimization plan (Phase 1 complete)

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* feat: add Codex session-start gate script with 4 verification gates (Phase 2 complete)

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* Changes before error encountered

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>
Co-authored-by: Richard Murillo <richard.murillo@example.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

* docs: standardize YAML array format for cross-platform compatibility (#923)

* Initial plan

* refactor: convert frontmatter to block-style YAML arrays in prompt and command files

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* docs: update frontmatter examples to use block-style YAML arrays

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* docs(governance): add YAML frontmatter array format constraint

- Add YAML Frontmatter Constraints section to PROJECT-CONSTRAINTS.md
- Include rationale with evidence from Session 826 RCA and GitHub Copilot CLI Issue #694
- Add validation checklist item for frontmatter arrays
- Add frontmatter validation requirement to SKILL-CREATION-CRITERIA.md
- Create session log for session 02

Refs: #898, Session 826

* docs: add issue URLs to YAML array format references

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>
Co-authored-by: Richard Murillo <richard.murillo@example.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

* docs: improve autonomous-issue-development.md structure (#566)

* docs: improve autonomous-issue-development.md structure

Expand documentation from 46 to 441 lines to match autonomous-pr-monitor.md style:

- Add "Common Development Patterns" section (5 validated patterns)
- Add "Troubleshooting" section (5 common scenarios)
- Enhance "Example Session Output" with TodoWrite and agent handoffs
- Add "Workflow Phases" table for quick reference
- Add "Agent Responsibilities" reference table
- Add "Prerequisites" and "Related Documentation" sections

Closes #506

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(session): add protocol compliance sections

Added Session Start and Session End checklist tables to match
the required session protocol format.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(security): prevent command injection in PR creation example

Addresses gemini-code-assist[bot] security review comment #2653014226.

The example showed a hardcoded PR title which could enable command
injection if autonomous agents construct titles from untrusted GitHub
issue titles containing shell metacharacters (e.g., $(reboot)).

Changes:
- Add security warning comment explaining CWE-78 risk
- Use read -r with process substitution to safely read issue title
- Demonstrate secure pattern for handling untrusted external input

This aligns with security principle of validating all external inputs.

Comment-ID: 2653014226

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: rjmurillo[bot] <rjmurillo-bot@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Claude <claude@anthropic.com>

* feat: implement local GitHub Actions validation and testing infrastructure (#925)

* Initial plan

* feat: Add actionlint validation for GitHub Actions workflows (Phase 1)

- Add actionlint to pre-commit hook (.githooks/pre-commit)
- Integrate actionlint into unified validation runner (scripts/Validate-PrePR.ps1)
- Add actionlint to Worktrunk pre-merge hooks (.config/wt.toml)
- Update SHIFT-LEFT.md documentation with workflow YAML validation section
- Update validation sequence to include step 3.5 (Workflow YAML)
- Create session log 2026-01-15-session-01-github-actions-local-testing.json

This implements Phase 1 (P0) of the GitHub Actions local testing epic.
actionlint catches 80%+ workflow YAML errors before push, reducing expensive CI iteration cycles.

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* feat: Add act support for local workflow testing (Phase 2)

- Create .actrc configuration with optimized defaults for act
- Create scripts/Test-WorkflowLocally.ps1 PowerShell wrapper for act
- Update SHIFT-LEFT.md with comprehensive local workflow testing documentation
- Document supported workflows (pester-tests, validate-paths, memory-validation)
- Document unsupported workflows (AI-dependent: ai-session-protocol, etc.)
- Add troubleshooting guide for act and PowerShell issues

This implements Phase 2 (P1) of the GitHub Actions local testing epic.
Enables local testing for PowerShell-only workflows without CI round-trips.

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* feat: Add yamllint for YAML style validation (Phase 3)

- Create .yamllint.yml configuration with GitHub Actions-optimized rules
- Add yamllint validation to scripts/Validate-PrePR.ps1 (validation 3.9)
- Add yamllint to pre-commit hook (.githooks/pre-commit) as non-blocking warnings
- Update SHIFT-LEFT.md with YAML style validation documentation
- Document yamllint vs actionlint relationship (style vs semantics)
- Configure yamllint as non-blocking to avoid impeding development velocity

This implements Phase 3 (P2) of the GitHub Actions local testing epic.
yamllint enforces YAML style consistency while actionlint handles semantic validation.

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* docs: Update session log with completion status

* fix: Address PR feedback on local testing infrastructure

- Add gh act extension to installation instructions
- Include .github/actions in actionlint validation scope
- Remove chocolatey recommendation, use GitHub releases instead
- Clarify Copilot CLI installation vs BOT_PAT requirement for AI workflows
- Switch to full production images in .actrc for maximum compatibility
- Update documentation to reflect full image usage and rationale

Addresses comments: 2693400619, 2693406075, 2693409506, 2693411769, 2693418268

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* refactor: Move Test-WorkflowLocally.ps1 to .claude/skills/github/scripts

- Move Test-WorkflowLocally.ps1 from scripts/ to .claude/skills/github/scripts/
- Update all documentation references in SHIFT-LEFT.md
- Update session log to reflect new location

Addresses comment 2699896328

Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>

* fix: address all PR #925 review comments

Fixes all 15 unresolved review threads from owner and automated reviewers:

1. Remove duplicate actionlint validation block in pre-commit hook
   - Kept newer comprehensive version (validates workflows + actions)
   - Removed old TIER 1 block (lines 269-313)

2. Fix wt.toml to validate both workflows and actions directories
   - Changed: actionlint .github/workflows/*.yml .github/actions/*.yml
   - Matches pre-commit hook validation scope

3. Remove Windows-specific chocolatey references
   - Prioritize cross-platform gh extension
   - Show macOS brew as alternative
   - Link to releases for manual download

4. Rename Write-Warning to Write-WarningMessage
   - Avoids shadowing PowerShell built-in cmdlet
   - Updated all call sites

5. Fix Validate-PrePR.ps1 to match both .yml and .yaml files
   - Changed -Filter to -Include with both extensions
   - Applies to both workflow and YAML validation sections

6. Fix session log field name inconsistencies
   - Changed handoffNotUpdated.evidence -> Evidence
   - Changed handoffNotUpdated.complete -> Complete
   - Matches other compliance entries

7. Add endingCommit to session log
   - Set to 38217dc (latest commit)
   - Required field per session protocol

8. Fix Test-WorkflowLocally.ps1 null handling
   - Check for null before Test-Path call
   - Prevents crash under Set-StrictMode when workflow not found

All fixes maintain existing functionality while addressing review feedback.

Related: #925

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: rjmurillo <6811113+rjmurillo@users.noreply.github.com>
Co-authored-by: Richard Murillo <richard.murillo@example.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

* fix: ADR-017 memory validation compliance (skill- prefix removal + bundled skill splitting) (#966)

* fix(pr-964): clean up session files and remove skill- prefix violations

Addresses two validation failures in PR #964:

1. Session file cleanup:
   - Moved 341 .md session files to .agents/archive/sessions/
   - Deleted 1 session file already in archive
   - All moved files have corresponding JSON equivalents

2. Memory file ADR-017 compliance:
   - Removed 87 duplicate memory files with skill- prefix
   - Renamed 56 unique memory files to remove skill- prefix
   - Zero skill- prefixed files remain
   - Added skills-serena-index to memory-index.md

These legacy files predate ADR-017 naming conventions and were
recovered from 52 orphaned branches. The skill- prefixed duplicates
existed alongside non-prefixed versions and have been cleaned up.

Note: Bundled skill validation failures are pre-existing in PR #964
and require separate handling via skill unbundling process.

Script: scripts/Fix-PR964-Validation.ps1

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* refactor: split 37 bundled skill files into 135 atomic files per ADR-017

Split bundled skill files to comply with one-skill-per-file standard.

Changes:
- Created scripts/Split-BundledSkills.ps1 automation script
- Split 37 bundled files containing 135 skills total
- Generated 110 new atomic skill files with proper naming
- Modified 25 existing skill files
- Deleted 37 bundled files

Validation:
- Validate-SkillFormat.ps1: PASSED
- markdownlint: PASSED (0 errors)

Context: PR #964 recovered orphaned files that predate ADR-017
Note: Memory index will be updated in a follow-up commit

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* docs: update session log to JSON schema v1.0

Update session-04 log to use new JSON schema format with protocolCompliance
and learnings sections per SESSION-PROTOCOL.md requirements.

Validation: Validate-SessionJson.ps1 PASSED

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* fix: convert skills-serena-index to table-only format per ADR-017

---------

Co-authored-by: rjmurillo-bot <noreply@github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

* fix(ci): use PATH export instead of source env for uv

The uv installer no longer creates an env file to source. Replace
`source $HOME/.local/bin/env` with `export PATH="$HOME/.local/bin:$PATH"`
to properly add uv to PATH in GitHub Actions workflows.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(ci): add --system flag to uv pip install

uv pip now requires either a virtual environment or the --system flag
to install packages. Add --system flag for GitHub Actions workflows
where we want to install directly to the system Python.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* feat(ci): enable Python tooling for agent contributions

- Add Python setup to setup-code-env composite action with:
  - enable-python and python-version inputs
  - Python version output
  - Python dependency installation via uv
  - Verification of ruff and pytest availability
- Enable Python 3.12 in copilot-setup-steps workflow
- Add Python dependency installation to bootstrap-vm.sh

This enables agents to contribute Python code with proper
tooling (ruff, pytest) available in the development environment.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(ci): use pip-audit without --requirement flag

The --requirement flag expects requirements.txt format, not pyproject.toml.
Running pip-audit without arguments audits installed packages instead.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(tests): patch SAFE_BASE_DIR for temp directory tests

- Monkeypatch _PROJECT_ROOT in validate_session_json tests
- Monkeypatch SAFE_BASE_DIR in invoke_skill_learning tests
- Fix tests checking 'extracted_learning' to use 'source' key

The path validation correctly rejects temp directories outside
project root. Tests now patch the base directory to allow temp
paths during testing while maintaining security in production.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(hooks): rename test_skill_context to check_skill_context

Pytest was collecting the function as a test because it started
with 'test_'. Renamed to 'check_skill_context' to prevent pytest
from treating it as a test function.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* docs: add test exit code interpretation as blocking constraint

- Add testing-exit-code-interpretation memory documenting that pytest
  "X passed, Y errors" output means test suite FAILED (non-zero exit)
- Update AGENTS.md Testing section with BLOCKING Test Exit Code
  Interpretation subsection
- Update CRITICAL-CONTEXT.md with explicit test exit code requirement
- Update memory-index with new memory for discoverability

Learning: "error" and "failed" are both non-pass outcomes in pytest.
Both result in non-zero exit code and must block commits.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Signed-off-by: Richard Murillo <6811113+rjmurillo@users.noreply.github.com>
Co-authored-by: Test <test@test.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Richard Murillo <richard.murillo@example.com>
Co-authored-by: Richard Murillo <6811113+rjmurillo@users.noreply.github.com>
Co-authored-by: rjmurillo-bot <rjmurillo-bot@users.noreply.github.com>
Co-authored-by: rjmurillo-bot <noreply@github.com>
Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Claude <claude@anthropic.com>