Curate the agent-maintained blog section (blog.tomrochette.com-content) by following the Daily refresh procedure in agents/AGENTS.md, pulling, working the article queue, updating and fact-checking articles, verifying links, appending to agents/log.md, and committing scoped to agents/. Use when the user says…
Analyze git history to identify high-churn files and suggest targeted improvements such as refactoring, test coverage, dead code removal, or library replacements.
Analyze a set of code changes (diff or files) and produce a structured test coverage report with three parts: introduced tests, change coverage, and uncovered code. Called by review-pr, review-implementation, and verify-pr; can also be invoked directly on any diff.
Downloads an arXiv HTML article, converts it to markdown with pandoc, archives it, and returns a structured summary. Use when processing a single arXiv paper given its HTML URL (e.g., https://arxiv.org/html/2504.12345). Called by the arxiv-catchup skill but can also be invoked directly.
Fetches new arXiv cs.AI papers published since the last processed date, processes each paper's HTML version via the arxiv-article skill, and advances the date checkpoint. Use when the user says /arxiv-catchup, catch up on arXiv, or wants to read recent AI papers.
Audit how your time splits between compounding and depreciating activities using the two-year test. Use when the user says /audit-attention, wants to check attention allocation, asks "where is my time going", or wants to know what to delegate vs. protect.
Audit the codebase for compatibility, the ISO/IEC 25010 characteristic covering co-existence (running alongside other software without conflict) and interoperability (exchanging data via standards). Finds hardcoded shared resources, dependency conflicts, missing API versioning, non-standard data formats, and brittle…
Audit project dependencies for outdated versions, known vulnerabilities, unmaintained packages, and license issues, then produce a prioritized upgrade and replacement plan.
Audit whether the software actually does what it should, completely and correctly. This is the Functional suitability characteristic of the ISO/IEC 25010 software quality model (completeness, correctness, appropriateness). Cross-checks implemented behavior against requirements in .sdlc/ and surfaces stubs, TODOs…
Audit the codebase for maintainability, the ISO/IEC 25010 characteristic covering modularity, reusability, analyzability, modifiability, and testability. Computes architectural metrics (coupling, fan-out, circular dependencies, layering violations, God modules) that the find- family does not cover, and aggregates…
Audit the codebase for performance efficiency, the ISO/IEC 25010 characteristic covering time behavior, resource utilization, and capacity. Finds N+1 queries, missing indexes, unbounded operations, blocking I/O in hot paths, algorithmic complexity, and missing backpressure. Use when the user says…
Audit the codebase for portability, the ISO/IEC 25010 characteristic covering adaptability (runs in different environments), installability (easy to install), and replaceability (easy to swap out or migrate off). Finds hardcoded paths/hosts/OS assumptions, config baked into code (violating 12-factor), vendor lock-in…
Scan the codebase for code-level security vulnerabilities including hardcoded secrets, injection risks, missing auth checks, and insecure patterns, then produce a prioritized remediation plan.
Audit the software's interfaces (CLI, API, and web UI) for usability, the ISO/IEC 25010 characteristic covering recognizability, learnability, operability, user error protection, and accessibility. Finds missing help text, inconsistent flags, poor error messages, missing confirmations on destructive actions, and…
Analyze what was done in the current session and surface concrete ways the workflow could have been automated to remove the user from the loop. Use when the user asks /automate-session, "how could this be automated?", or "what could we automate from this session?".
Walk the SDLC artifact chain in reverse (code back to issue) to verify end-to-end consistency and traceability. Use when the user says /backpropagate-sdlc, "backpropagate", "verify SDLC trace", or wants to detect drift accumulated across artifacts after the forward pipeline has run.
Determine whether a GitHub issue is already addressed in the codebase by extracting its requested behavior and verifying it against the code. Works for feature requests and bug reports in any format, with or without formal acceptance criteria. Use before starting work on an issue, or to triage stale issues, to avoid…
Orchestrate check-issue-status across a list of issues to find which are already addressed in the code. Supports checking every open issue in a repository, issues authored by a specific user, or issues matching a custom search query, and these scopes combine. Use to triage a backlog for stale or already-implemented…
Detect pull requests someone else linked to the issue being worked on, then offer to continue, stop, or review them. Use mid-flow to avoid duplicating in-progress work and to decide whether to depend on an external PR instead.
Compare skills between two directories (mine and theirs) to identify best practices from "theirs" that can be adopted into "mine". Produces a prioritized list of improvements with concrete suggestions.