SR&ED Audit-Proof Commit Message Writing (Claude Code memory tested)

Source: Notion | Last edited: 2025-07-25 | ID: 23b2d2dc-3ef...

---

APCF: Audit-Proof Commit Format for SR&ED Evidence Generation

• First get the current ‘America/Vancouver’ time using:

TZ='America/Vancouver' date "+%A %Y-%m-%d %H:%M:%S %Z %z"

Tool Usage Preferences

• File Operations: Prefer Read, LS, Glob, Grep over MCP filesystem tools (broader access)

Usage: Claude Code Interaction

When you request “APCF”, I will analyze ALL changes:

• Staged files (git diff --cached) - ready to commit
• Modified files (git diff) - unstaged changes
• Untracked files (git status --porcelain) - new files
• Auto-derive SR&ED evidence from complete change analysis
• Generate logical commit grouping and sequencing strategy
• Create audit-proof commit messages for each logical group Workflow: Make changes → Request “APCF” → I analyze everything → Suggest commit strategy

Format Template

• Each commit in the sequence uses this format: auto-generated from workspace analysis
• Quote in backticks (`) any technical noun or proper noun per Markdown/GitHub style: package, library, tool, command, file, directory, language, class, function, config key, or other technical term; do not quote pronouns or non-technical nouns.

type(scope): description

- Knowledge Gap: [Auto-derived from file patterns + technical domain uncertainty + failed approaches]
- Motivation: [Auto-derived from commit intent + workspace changes + timeline context]
- Hypothesis: [Auto-derived from commit intent + proposed technical approach + risk factors]
- Investigation: [Auto-derived from workspace analysis + systematic methodology + failures/iterations]
- Result: [Auto-derived from changes + technical advancement + specific measurements]
- Authenticity: [Developer notes + work timestamps + debugging context for CRA contemporaneous compliance]

The follow footer section display the lines of libraries involved seperated by commas and spaces. The lines are shown only if the pertaining libraries are involved:

- PyOpen: {Publicly available third-party Python libraries (on PyPI)}
- PyPriv: {Private or internal Python libraries not on PyPI}
- PyOthr: {Third-party programming libraries that are not Python, e.g. C++, JavaScript, Java, etc.}

Here in this line, the last line in the commit message, we display the result of the current 'America/Vancouver' time.

Commit Grouping Logic

Logical Sequencing Strategy:

1. Emergency First (hotfix:, revert:) - Critical fixes and risk mitigation
2. Infrastructure First (build:, config:, deps:, ci:) - Foundation changes
3. Core Implementation (feat:, refactor:, perf:) - Main functionality
4. Quality Assurance (test:, fix:, security:) - Validation and corrections
5. Documentation (docs:) - Knowledge capture
6. Release Management (release:) - Deployment readiness
7. Maintenance (style:, chore:) - Process improvements
8. Work in Progress (wip:) - Development snapshots (avoid in production) Atomic Grouping Rules:

• Related files together - Files that implement the same feature
• Dependency respect - Infrastructure before features that depend on it
• Audit trail clarity - Each commit tells complete SR&ED story
• Rollback safety - Each commit is independently functional

Auto-Derivation Intelligence (Workspace State → SR&ED Evidence)

File Pattern → Domain Detection

• .py, *.ipynb → Algorithm/ML Development
• .js, *.ts, *.jsx → Frontend/API Innovation
• .sql, *.db → Database Architecture Research
• .yaml, *.json, *.toml → Configuration Investigation
• test_*, *.test.* → Validation Methodology
• Dockerfile, *.sh → Infrastructure Innovation
• .md, docs/ → Knowledge Capture Investigation

Workspace Analysis → SR&ED Scope & Priority

• Single file change → Focused technical uncertainty
• Multiple related files → Comprehensive investigation
• Cross-domain changes → System-wide innovation
• New file additions → Experimental development
• Dependency changes → Technology integration research

APCF Evidence Standards

• Specificity: Use actual counts, technology names, file types (“modified 3 files” not “comprehensive changes”)
• Facts over Interpretation: State direct technical actions (“what was built” not “how well it performs”)
• CRA Compliance: Include failure documentation, work-commit timestamps for contemporaneous evidence
• Avoid derived metrics: No “efficiency ratios”, “performance improvements”, or calculated benefits

Search & Audit Trail Integration

• Every commit becomes searchable audit evidence
• Cross-repository SR&ED pattern recognition
• Automatic evidence chain building for quarterly reports
• Government audit trail with direct commit verification