ai-tpk

Create new AI agents or modify existing agents in this repository. Use when users want to create a new specialized agent, update an existing agent definition, add agent documentation, or understand the agent creation process.

Guide for sensible file and module organization. Use when designing file structure for a new feature, deciding whether to split or merge files, refactoring a large file, evaluating whether a new file is justified, deciding whether to create a new directory, choosing where test files should live, or naming a new directory.

Methodology for detecting documentation gaps in a codebase, used during gap-analysis phases to produce a structured, severity-tiered list of missing or inaccurate user-facing documentation.

REQUIRED for all pull requests. Always use when opening PRs, creating merge requests, or executing gh pr create / glab mr create commands. Enforces conventional branch names and conventional-commit PR titles.

Guide for writing reliable, deterministic automated tests. Use when writing new tests for a feature or bug fix, reviewing or refactoring existing test code, creating setup and teardown logic, choosing between unit/integration/e2e testing, designing test fixtures or factories, troubleshooting flaky tests, or writing mocks and stubs. Apply this skill whenever test code is being authored or evaluated, even if the user does not explicitly ask for test quality guidance.

Defines the high-signal review contract for code and diff reviews. Use whenever the user asks for a code review, PR review, diff review, asks to review a pull request, asks "is this code OK to merge", asks to identify issues in a changeset, or invokes a `/review` command — even when the user does not explicitly say "skill". This skill determines what qualifies as a reportable finding, what does not, and how findings are structured and validated before being reported. Invoke it before surfacing any review findings, not after.

Capture lightweight observations about friction, surprises, or repeated mistakes during a Claude Code session, so they can be reviewed later and turned into skill improvements. Use whenever you notice a pattern worth remembering — confusing instructions in a skill, a workaround you had to invent, a step that should be automated, or behaviour that contradicted a skill's guidance. Always log when you correct yourself or the user corrects you on something a skill should have prevented.

REQUIRED before opening any pull request. Always invoke when about to open a PR, create a merge request, push for review, or run pre-PR checks. Runs lint and format checks as a gate before PR creation.

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

Defensive coding standards for bash scripts authored as deliverables. Apply when writing or editing any `.sh` file or a script whose first line is a bash shebang — even when the user does not ask for best practices explicitly.

Guide for creating well-structured conventional commits. Use when creating git commits, drafting commit messages, or reviewing commit message quality.