📄 CodeRabbit inference engine (.agents/governance/interviews/security-interview.md)

Review infrastructure code changes including GitHub Actions workflows, git hooks, and Dockerfile for security vulnerabilities

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/metrics/baseline-report.md)

Infrastructure files (including .githooks/pre-commit) must receive security review before deployment

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/metrics/dashboard-template.md)

Ensure infrastructure files in GitHub workflows directory are reviewed by agents with target rate of 100%

Implement full CodeRabbit enforcement for CI/CD workflow files in .github/workflows/ directory

Detect GitHub Actions workflow files (.github/workflows/*) as infrastructure changes requiring security and devops agent review

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/security/infrastructure-file-patterns.md)

GitHub Actions workflow files (.github/workflows/*.yml and .github/workflows/*.yaml) should trigger security agent review due to critical security implications

.github/workflows/*.{yml,yaml}: Infrastructure files (.github/workflows/) require devops and security agent review before commit
Add pre-commit validation that fails on infrastructure files (.github/workflows/) lacking security review documentation
Infrastructure files must include a security impact analysis document before creation or modification
Retroactive security review is required for quick-fix changes to workflow files made without pre-commit review
Workflow files must validate that all required permissions are explicitly declared and justified

Keep workflow YAML files light and thin - MUST NOT put complex logic in workflow YAML files; use thin workflows with calls to external testable modules or skills

GitHub Actions workflows should remain thin and delegate complex logic to PowerShell modules rather than embedding bash-like commands in workflow YAML run blocks

.github/workflows/*.{yml,yaml}: Never use || true pattern without explicit error handling and ADR justification
Do not use || true error suppression in workflows without documenting rationale in an ADR

Use thin workflow pattern: minimize logic in YAML files, delegate complex operations to PowerShell modules

.github/workflows/*.{yml,yaml}: Use YAML heredoc with proper indentation for multi-line GitHub Actions values
Use GH_TOKEN for auto-authentication in GitHub Actions shell scripts
Use fixed-length lookbehinds for regex patterns in GitHub Actions
GitHub Actions outputs must be single-line format
Matrix jobs in GitHub Actions should use artifacts for output storage
Use environment variables for shell command interpolation in GitHub Actions to prevent injection vulnerabilities
Use verdict tokens for structured AI output in GitHub Actions
Format PR comments following CodeRabbit-style structured comment conventions

Implemen...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/analysis/pr41-issue-analysis.md)

.github/workflows/*.yml: Use dotnet-nunit reporter in dorny/test-reporter@v1 when Pester tests output NUnitXml format
Add explicit permissions block to GitHub Actions workflow jobs, specifying minimal required permissions (e.g., contents: read and checks: write for test reporting workflows)

.github/workflows/*.yml: Use minimal GitHub Actions permissions in CI workflows
Add workflow_dispatch trigger to GitHub Actions workflows for manual triggering capability
Add PowerShell-Yaml module installation step in CI workflows
Use ubuntu-latest for drift detection CI workflows for lighter weight execution
Add issue deduplication for drift alerts in GitHub Actions workflows
Use path filtering appropriately scoped in GitHub Actions workflows

CI/CD pipeline must include a drift-linting job that compares Claude agent content against generated VS Code and Copilot CLI variants on every PR modifying agent files

Create automated CI validation to detect absolute paths (Windows, macOS, Linux patterns) in markdown files with colored output, line numbers, and remediation steps

Configure concurrency groups in GitHub Actions workflows to prevent comment spam and race conditions when multiple commits trigger parallel workflow runs on the same PR

.github/workflows/*.yml: Remove silent || true error suppression from critical operations; instead track failures in arrays and provide summarized error reporting at workflow completion
Use minimal token scopes in GitHub Actions workflows: use repo scope for read-only operations, issues:write for issue modifications, and workflow permissions declarations following principle of least privilege

.github/workflows/*.yml: Store heredoc content in separate files instead of inline within YAML run: blocks to avoid YAML parsing errors when content lacks proper indentation
Use github.token (workflow's built-in token) for PR comments in GitHub Actions instead of service account PATs, as it automatically has write access t...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/retrospective/pr-feedback-remediation.md)

Use 'ubuntu-latest' for CI runner preference instead of 'windows-latest' for significantly faster performance, except for PowerShell Desktop or Windows-specific features

.github/workflows/**/*.{yml,yaml}: Workflows should be thin and contain only orchestration logic; all business logic (parsing, validation, formatting) must be extracted to PowerShell modules (.psm1)
Keep workflow YAML files to a maximum of 100 lines (orchestration only)
Do not parse complex strings (verdict, labels, etc.) in workflow YAML run: blocks; delegate to PowerShell modules
Do not validate business rules directly in workflow YAML; delegate to PowerShell modules
All PowerShell scripting for workflows must be PowerShell-only (no bash or other shell languages embedded in workflows)

.github/workflows/**/*.{yml,yaml}: PowerShell used consistently for label and milestone parsing in AI workflow (extract parsing logic to AIReviewCommon.psm1::Get-LabelsFromAIOutput rather than inline bash)
Use ::error:: GitHub Actions format for error output instead of echo or Write-Error to ensure errors are properly captured and cause workflow failure
Remove all '|| true' patterns and silent failure suppression in workflow steps - errors must be aggregated and reported explicitly using ::error:: format
Add token masking with ::add-mask:: before using secrets in workflow run steps to prevent accidental token leakage in logs
Replace secrets.BOT_PAT with github.token for issue and PR operations to minimize token privileges and reduce security attack surface
Document minimum required token scopes in workflow comments for each operation (e.g., 'repo' scope for issue edit and label application, 'read' scope for readonly operations)
Extract AI workflow parsing logic to testable PowerShell module functions (AIReviewCommon.psm1) rather than inline bash - enables unit testing in isolation and reduces regression risk

.github/workflows/**/*.{yml,yaml}: Use bash/PowerShell error collection pattern: initialize FAILED_...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/skills/ci-runner-preference.md)

**/.github/workflows/*.{yml,yaml}: Use ubuntu-latest (or linux-latest) runners for GitHub Actions workflows instead of windows-latest for better CI/CD pipeline performance
Only use windows-latest runner when PowerShell Desktop (5.1) is required, Windows-specific features are needed, or testing Windows-only scenarios
Prefer PowerShell Core (pwsh) over PowerShell Desktop for GitHub Actions scripts to enable cross-platform compatibility on Linux runners

Workflows SHOULD be less than 100 lines, containing orchestration only. Business logic SHOULD be extracted to PowerShell modules (.psm1).

PowerShell workflow files must follow the thin-workflow pattern where orchestration logic stays in the workflow and business logic is extracted to reusable modules

Maintain consistent token scopes and authentication strategy throughout workflows; use the same token source for all related operations to avoid permission errors

**/.github/workflows/*.{yml,yaml}: Avoid code injection vulnerabilities by never directly interpolating unsanitized GitHub context variables (like ${{ github.event.issue.title }} or ${{ github.event.issue.body }}) into bash commands; instead save them to files using heredoc syntax and read them back
Use toJSON() to escape GitHub context variables when passing them to bash commands, or employ input validation before using variables in shell contexts
GitHub Actions workflows should request only the minimum necessary permissions (principle of least privilege) required for their operations

**/.github/workflows/*.{yml,yaml}: When ≥2 GitHub Actions workflows share logic, extract to composite action with parameterized inputs for reusability
GitHub Actions workflow files should include YAML syntax validation and actionlint checks in pre-commit configuration

**/.github/workflows/*.{yml,yaml}: Use artifacts instead of job outputs for passing data between matrix jobs in GitHub Actions workflows
Use environment variables for shell variable expans...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/steering/security-practices.md)

Implement OWASP Top 10 prevention practices in GitHub Actions workflows

Use secure defaults and minimal exposure for security misconfiguration prevention

Files:

• .github/workflows/pr-maintenance.yml

⚙️ CodeRabbit configuration file

.github/workflows/**: SECURITY-CRITICAL PATH - ASSERTIVE ENFORCEMENT
Review for: - Authentication and authorization (secrets handling, OIDC) - Credential exposure (env vars, git output, logs) - Injection vulnerabilities (command injection, script injection) - Race conditions (concurrent jobs, artifact operations) - Resource exhaustion (limits on runners, matrix expansions)
Flag: - Unquoted variables in shell commands - Direct use of user input in scripts - Missing error handling and early exit conditions - Hardcoded secrets or credentials - Unsafe use of 3rd party actions
DO NOT IGNORE any security-related findings on this path.

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/architecture/DESIGN-REVIEW-pr-60-remediation-architecture.md)

Use GitHub Actions workflow steps with PowerShell for JSON parsing and data manipulation; reserve bash only for simple npm install, git operations, and environment variable exports

Aggregate errors across workflow steps using error accumulation arrays and export failure summary via GITHUB_OUTPUT

.github/workflows/**/*.yml: GitHub Actions workflows orchestrating multiple agents should use matrix strategy with fail-fast: false to ensure all reviews complete
Use artifact passing (not job outputs) to transfer data between workflow jobs when using matrix strategy
Implement idempotency in workflow comment generation using comment markers to distinguish update vs create operations
Use skip optimization to check for docs-only changes before executing expensive AI review jobs
Add CI validation step to verify prompt templates include required output format markers before merging

.github/workflows/**/*.yml: Replace bash parsing with PowerShell functions in GitHub Actions workflows - use Get-LabelsFromAIOutput and Get-MilestoneFromAIOutput from AIReviewCommon.psm1 instead of bash grep/tr/xargs to prevent command injection (CWE-78) and word splitting vulnerabilities (CWE-20)
In GitHub Actions workflows parsing AI output, never use bash xargs, unquoted variable expansion, or for loops with simple string parsing - replace with PowerShell cmdlets that handle structured output safely
Verify PowerShell Core is available on GitHub Actions ubuntu-latest runner before using PowerShell in workflows

Add checkout step to skip-validation job in GitHub Actions workflows that use dorny/paths-filter

.github/workflows/**/*.yml: Add explicit exit 0 to PowerShell scripts in GitHub Actions workflow run blocks to prevent $LASTEXITCODE persistence
Validate GitHub labels exist before deploying workflows that reference them with gh pr edit --add-label
Add cross-platform validation matrix to test workflows or document single-platform requirements upfront when tests have platf...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/governance/PROJECT-CONSTRAINTS.md)

**/.github/workflows/**/*.{yml,yaml}: MUST NOT put business logic in workflow YAML files
SHOULD keep workflows under 100 lines (orchestration only)

Workflows must be thin and composed of testable, reusable modules

Follow Copilot follow-up PR naming convention: copilot/sub-pr-{original_pr_number} for branches targeting the original PR's base branch

**/.github/workflows/**/*.{yml,yaml}: Implement concurrency control using GitHub Actions concurrency group to prevent simultaneous PR maintenance runs
Configure GitHub Actions workflows to create an issue on failure with workflow diagnostics
Design thin GitHub Actions workflows that orchestrate logic rather than implementing testable module logic directly

**/.github/workflows/**/*.{yml,yaml}: Add workflow action steps AFTER parsing to apply suggested labels, assign suggested users, post feature review as comment, and update issue milestone based on recommendation
Specify exact line numbers and file locations in rollback plan, including which lines to remove from .github/workflows/ai-issue-triage.yml and triage summary PowerShell script
Include validation logic in workflow to verify that issue labels and analyst categorization match (e.g., if labeled 'enhancement', categorized as 'enhancement')
Specify Copilot model parameter in feature review workflow step (use claude-opus-4.5 for consistency with PRD generation step)
Define error handling for parse failures in workflow: specify fallback behavior when regex extraction fails (e.g., if recommendation returns UNKNOWN)
Verify category detection accuracy by testing whether analyst agent reliably categorizes feature requests, including false negative rate and fallback behavior for miscategorization

All workflow jobs must include explicit copilot-model parameter specification per governance requirements

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/planning/PR-60/001-pr-60-review-gap-analysis.md)

Use unique identifiers (e.g., run ID, timestamp) for temporary file paths in CI/CD workflows to prevent conflicts in concurrent runs on shared runners

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/planning/PR-60/004-pr-60-implementation-review.md)

**/.github/workflows/**/*.yml: Replace bash label parsing in GitHub Actions workflows with PowerShell validation to prevent command injection via untrusted AI output
Validate all label strings against regex pattern '^[\w-.\s]{1,50}$' before applying to GitHub issues
Validate all category strings against regex pattern '^[\w-]{1,30}$' before using in workflow outputs
Validate all milestone strings against regex pattern '^[\w-.\s]{1,50}$' before assigning to issues
Use explicit error handling with informative error messages instead of '|| true' for GitHub CLI (gh) commands that should not fail silently
Keep '|| true' only for commands where empty/missing results are valid expected outcomes (e.g., grep with no matches, find with no results)
Verify GitHub CLI authentication status is successful before attempting API operations in CI workflows
Log validation rejections with details (invalid format, rejected label) for debugging label application failures
Provide detailed error messages in GitHub Actions workflow steps including context (what failed, why, what to check)
Use PowerShell 'Write-Host' with color output for structured workflow logging instead of plain echo statements
Use '|| true' in workflows only for commands where missing data is an expected valid state; document the reason in comments
Validate workflow inputs and outputs against strict regex patterns before using them in GitHub API commands

**/.github/workflows/**/*.yml: Use environment variables instead of direct ${{ }} template syntax in shell scripts to prevent shell injection attacks
Use artifacts instead of job outputs to pass data between matrix job legs, since outputs only expose one matrix leg
Prepare multi-line content via prior steps with GITHUB_OUTPUT heredoc syntax rather than attempting to execute shell commands in YAML with: inputs

**/.github/workflows/**/*.yml: Add explicit 'exit 0' at end of PowerShell workflow scripts to prevent $LASTEXITCODE persistence
Validate GitHub lab...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/planning/PR-60/007-phase-1-detailed-schedule.md)

Use PowerShell-only code (no bash, Python, or shell scripts) with all workflow steps configured with 'shell: pwsh' per ADR-006

Use quoted heredoc (with 'EOF' delimiter) for safely handling github.event.pull_request.title and github.event.pull_request.body in bash env vars to prevent code injection

GitHub Actions workflows must use PowerShell functions from exported modules rather than inline bash parsing (grep/sed/tr) for consistent security and testability

In workflow comments, use explicit language about verdict behavior and consequences (e.g., 'should cause the check to FAIL (block merge)' rather than ambiguous phrasing)

**/.github/workflows/*.yml: Use PowerShell syntax (shell: pwsh) for GitHub workflow steps when handling untrusted AI output
Import PowerShell modules using correct relative path format (Import-Module .github/scripts/AIReviewCommon.psm1) in GitHub workflows to ensure module dependencies are properly resolved
Use PowerShell environment variables for all user input in GitHub workflows instead of direct bash variable interpolation to eliminate word-splitting vulnerabilities

Use $env:GITHUB_WORKSPACE with absolute paths instead of relative ./ prefix when importing PowerShell modules in GitHub Actions workflows

Include env: GH_TOKEN in workflow steps that use GitHub CLI to ensure proper authentication

**/.github/workflows/*.yml: GitHub Actions workflow JSON output consumption must use fromJson() to parse matrix data from upstream job outputs
GitHub Actions workflow matrix strategy should include max-parallel constraint to throttle concurrent job execution and respect API rate limits

Document all CI/CD workflow changes for backward compatibility verification in workflow recommendation documents

**/.github/workflows/*.yml: Add workflow input parameter to toggle alignment validation on/off for rollback scenarios without breaking CI
Embedding validation should run as separate independent workflow step, not integrated into Valid...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/planning/pr-60-focused-plan.md)

Use sed with POSIX character classes instead of grep -P for regex patterns to ensure portability across Linux and macOS (grep -P is GNU-only)

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/planning/pr-60-security-review.md)

.github/{workflows,scripts}/**/*.{yml,yaml,sh}: Avoid leaking sensitive information or internal configuration details in error messages and workflow logs that could be visible to attackers. Use generic error messages in production and reserve detailed diagnostics for debug logs only.
Sanitize and validate all untrusted user input (such as GitHub event data like PR titles, issue bodies, commit messages) before using in shell commands, API calls, or other operations that could interpret special characters.
When updating GitHub comments in workflows, use idempotent markers (HTML comments with unique identifiers) in comment bodies to enable safe updates by specific comment ID rather than relying on comment position.

Enforce ADR-005 with pre-commit hook rejecting bash in .github/workflows/ and .github/scripts/ directories

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/retrospective/2025-12-18-ai-workflow-implementation.md)

AI automation in CI/CD requires verdict tokens (PASS/WARN/CRITICAL_FAIL) for deterministic bash parsing without AI interpretation

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-18-session-04-ai-workflow-debugging.md)

Avoid explicit gh auth login --with-token when GH_TOKEN environment variable is already set, as gh CLI automatically uses the env var and will fail with exit code 1

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-18-session-13-workflow-lessons.md)

Use composite actions to encapsulate reusable workflow steps (Node.js/npm/Copilot CLI setup, authentication, agent loading, verdict parsing)

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-18-session-13-workflow-lessons.md)

Convert single-job loops to multi-job matrix structures (detect → validate → aggregate) for parallel validation with artifact passing

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-18-session-34-pr-60-comment-response.md)

{.github/**/*.{sh,yml,yaml},**/*.sh}: Avoid gh auth login authentication patterns; use session-based authentication instead (reference Session 04 for implementation details)
Use secret masking for sensitive tokens like BOT_PAT in shell commands and CI/CD workflows

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/qa/057-pr-222-import-module-standardization.md)

Standardize Import-Module paths to use $env:GITHUB_WORKSPACE variable instead of relative paths

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-21-session-58-pr-141.md)

All skip jobs in workflows that document a checkout requirement must include the checkout step

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/devops/pr-automation-script-review.md)

.github/workflows/pr-maintenance.yml: Use GitHub Actions scheduled workflow (cron: '0 * * * ') for hourly PR maintenance automation instead of Windows Task Scheduler or systemd timers
Implement concurrency control using GitHub Actions concurrency group to prevent simultaneous workflow runs: concurrency { group: pr-maintenance, cancel-in-progress: false }
Check GitHub API rate limits before PR maintenance execution and fail with error if remaining calls < 200 per resource (core: 100, search: 15, graphql: 100, code_search: 5)
Use environment variables for configuration (PR_PROTECTED_BRANCHES, PR_BOT_AUTHORS, PR_ACK_REACTION, PR_MAX_PRS) with sensible defaults to avoid hardcoded values
Restrict GitHub Actions job permissions to minimum required: contents: read, pull-requests: write, issues: write
Post workflow summary to GitHub Actions step summary ($GITHUB_STEP_SUMMARY) including metrics table, rate limit status, and blocked PR list with workflow run link
Create automatic alert issue on workflow failure with labels [automation, P1] and link to failed workflow run for visibility and tracking
Create automatic alert issue when blocked PRs are detected (require human action) with labels [automation, needs-triage] and include blocked PR list and workflow run link
GitHub Actions workflow timeout should be 10 minutes (timeout-minutes: 10) to prevent runaway jobs while allowing buffer above expected <2min runtime for 20 PRs
Upload workflow logs as artifacts with 30-day retention (actions/upload-artifact@v4) to .agents/logs/pr-maintenance.log for historical audit trail
Use ubuntu-latest GitHub Actions runner (not Windows) because it provides PowerShell Core 7.4+ pre-installed and avoids Windows-specific dependencies
Always run security-critical steps (environment validation, rate limit check, secret verification) before PR maintenance processing to fail fast
Use GitHub API rate limit bucket-specific thresholds: core=100 (2% of 5000), search=15 (50% of 30), graphql=100 (2% ...

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/retrospective/2025-12-22-pr-249-comprehensive-retrospective.md)

Parameterize branch references instead of hardcoding values like 'main' in workflow scripts and automation code

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-22-session-64-pr-automation-devops-review.md)

Schedule PR maintenance workflow to run hourly using GitHub Actions with performance target of <2 minutes for 20 PRs

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-22-session-67-pr-249-review-response.md)

{**/.github/workflows/*.yml,**/*.ps1}: Default dry_run to true when inputs.dry_run is empty to prevent DryRun bypass for scheduled GitHub Actions runs
Allow protected branch checks to pass when GITHUB_ACTIONS=true to prevent CI workflow failures

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-22-session-72-pr-249-retrospective.md)

**/.github/workflows/*.{yaml,yml}: When workflow inputs are empty (scheduled triggers), default to fail-safe mode (dry_run=true) instead of fail-open
Explicitly declare environment variables (secrets, tokens) in each GitHub Actions workflow step that needs them, rather than relying on job-level inheritance

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-22-session-79-pr-268-review-response.md)

Use unique identifiers (such as issue numbers) in test cleanup operations to prevent race conditions across concurrent workflow runs

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/architecture/ADR-014-github-actions-arm-runners.md)

Do not migrate Windows-based GitHub Actions workflows to ARM runners; keep Windows workflows on windows-latest as ARM runners are not available for Windows

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-23-session-83-adr-017-quantitative-analysis.md)

Create CI validation for index ↔ atomic file consistency when implementing tiered index expansion

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/critique/001-pr365-remediation-critique.md)

Include note about URL staleness in CI/CD documentation that references specific run IDs, recommending use of dynamic commands (e.g., gh run list --branch <branch-name>) to find latest runs

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/analysis/pr-quality-gate-320c2b3-analysis.md)

Disable fail-fast in matrix workflow jobs for PR processing to prevent single PR failures from blocking other PRs

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-23-session-85-pr-310-review.md)

Model routing for Copilot CLI must use evidence-aware, tiered routing: JSON extraction → gpt-5-mini, general review → claude-sonnet-4.5, security → claude-opus-4.5, code evidence → gpt-5.1-codex-max

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/sessions/2025-12-23-session-90-adr-debate-clarification.md)

Aggregator model routing must be enforced via branch protection rules, not just code-level checks

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/analysis/ADR-023-related-work-research.md)

Validate matrix-level aggregation of Quality Gate failures to prevent masking of genuine failures

Files:

• .github/workflows/pr-maintenance.yml

📄 CodeRabbit inference engine (.agents/governance/interview-response-template.md)

Primary deliverables from agents should be saved to .agents/[category]/[pattern].md with naming convention [PREFIX]-NNN-[description].md

Single-source agent files should use frontmatter markers to delineate platform-specific sections for VS Code and Copilot CLI variants

Maintain artifact synchronization markers in tracking files (.md) with status indicators ([COMPLETE], [RESOLVED], [VERIFIED]) and timestamps to document completion and verification of work

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/critique/001-agent-templating-critique.md)

For agent platform files, evaluate whether near-identical variants (99%+ overlap) can be consolidated with conditional configuration rather than maintaining separate files

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/retrospective/pr43-coderabbit-root-cause-analysis.md)

Include all verified facts with exact values (percentages, hours, counts) in escalation prompts to higher-level agents, rather than summarized or rounded values

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/retrospective/pr43-coderabbit-root-cause-analysis.md)

.agents/**/*.md: Use PREFIX-NNN naming convention (e.g., EPIC-001, CRITIQUE-001) for sequenced artifacts and type-prefixed naming (e.g., prd-, tasks-) for non-sequenced artifacts
Normalize all file paths in markdown documents to be repository-relative before committing, removing absolute machine-specific paths

.agents/**/*.md: Session logs and documentation must include Phase checklist verification (Phase 1-3 protocol compliance including agent activation, instruction reading, handoff file updates, and session logging)
Session logs must document Session ID, date, agent name, and branch information in a standardized header format

All artifact files in .agents/ must be in Markdown format

Document analysis recommendations with specific rationale when adding new governance documents like PROJECT-CONSTRAINTS.md

Maintain debugging skills documentation in .agents/ directory

Document implementation notes explaining deviations from user prompts or decisions made during development (e.g., using plural form for directory names)

Run markdown lint on all generated artifacts before completing a session

Run markdown lint validation (0 errors expected) before committing documentation files in the .agents directory

Planning artifacts in .agents/ directory must maintain effort estimate consistency within 20% threshold, ensure no orphan conditions exist (all linked to tasks), and complete requirement coverage

.agents/**/*.md: Use consistent absolute file paths throughout task and PRD documentation instead of mixing relative and absolute path formats
Run markdown linting with npx markdownlint-cli2 --fix on all agent-generated documentation before commit

All modifications to agent documentation and specifications must be marked with status updates (e.g., DRAFT → CONSOLIDATED) and include consolidation notes in headers

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/AGENTS.md)

Critique artifacts must follow naming pattern NNN-feature-critique.md

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/sessions/2025-12-23-session-62-adr-017-review.md)

.agents/critique/*.md: Document architecture review verdicts with clear APPROVED/REJECTED/APPROVED WITH CONDITIONS status and supporting evidence
Include comprehensive findings, key issues, and recommendations in architecture critiques

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/critique/001-pr365-remediation-critique.md)

Verify existence of referenced documentation files before updating them in automation procedures

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/handoffs/copilot/sub-pr-310/session-92-93.md)

Move debate logs to .agents/critique/ directory

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.github/copilot-instructions.md)

Store critique outputs in .agents/critique/ directory

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/qa/063-session-63-guardrails-critique-qa.md)

Critique documents should be stored in the .agents/critique/ directory with a numeric prefix following the format NNN--critique.md

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

📄 CodeRabbit inference engine (.agents/critique/465-spec-validation-false-positive.md)

Require explicit verdict patterns in all AI agent outputs rather than relying on substring keyword matching for verdict detection

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

⚙️ CodeRabbit configuration file

Agent configuration files. Only flag security issues or broken cross-references. Ignore style, formatting, and structure.

Files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

Learnt from: rjmurillo
Repo: rjmurillo/ai-agents PR: 60
File: .agents/critique/003-pr-60-remediation-critique.md:6-6
Timestamp: 2025-12-19T00:35:21.906Z
Learning: In planning, critique, and analysis documents under .agents/critique, do not include manual revision tracking (e.g., version numbers, revision history tables, or Last Updated dates). Rely on Git commit history for version tracking, including what changed, when, who, and why (documented in commit messages).

Applied to files:

• .agents/critique/001-issue-360-bot-loop-prevention-critique.md

[uncategorized] ~23-~23: The official name of this software platform is spelled with a capital “H”.
Context: ...on to the Process PR comments step in .github/workflows/pr-maintenance.yml. When a P...

(GITHUB)