using-agent-skills
Selects and applies the smallest relevant set of skills for a task, preventing generic prompting and unused specialist guidance. Use at the start of every command, task, review, or repair cycle.
- License 许可
- Apache-2.0
- 最近更新
- 2026-07-13
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- using-agent-skills
- description
- Selects and applies the smallest relevant set of skills for a task, preventing generic prompting and unused specialist guidance. Use at the start of every command, task, review, or repair cycle.
03 SKILL.md 正文
Using Agent Skills
Purpose
Ensure the agent deliberately activates the right skills instead of relying on a single broad instruction. The harness should be skill-routed, not vibes-routed.
Required workflow
- Identify the lifecycle stage: intent, context, expectations, task derivation, implementation, review, validation, commit, reporting, delivery, teaching.
- Select only the skills that materially affect the next action.
- Read each selected skill's trigger, constraints, red flags, and verification checklist.
- State the selected skills in the action record or result artifact.
- Do not invoke project-pack skills unless the configured project pack is active.
- Do not use a domain skill as a substitute for generic test, engineering, or validator review.
Skill routing baseline
| Stage | Required skills |
|---|---|
| Intent | intent-capture |
| Context | context-assembly, environment-bootstrap |
| Expectations | expectation-derivation, expectation-validation |
| Task derivation | epic-task-decomposition |
| Build | red-green-vertical-slice, simple-python-implementation, testing-patterns |
| Review | code-review, gilfoyle-code-review when high scepticism is needed |
| Python design | simple-python-implementation, oop-design-patterns only when OO is justified |
| Comments/docs | self-explanatory-code-comments |
| Containers | docker-containerization |
| CI/CD | github-actions-ci-cd |
| Azure IaC | azure-naming, bicep-best-practices |
| Commit | verified-commit |
| PR | pull-request-delivery |
| Session report | session-reporting |
| Teaching | teach-me |
Red flags
- A command proceeds without naming any active skill.
- The agent uses every skill, creating context noise.
- A specialised skill is used outside its domain.
- A reviewer uses implementation skills to modify code.
- The builder sees validator-only hidden checks.
Verification
- [ ] Skill selection is recorded.
- [ ] Selected skills match the lifecycle stage.
- [ ] Domain skills match the active project pack.
- [ ] No irrelevant high-token skill was loaded.
04 同仓库更多 Skills
invoice-domain-review
Domain review for contractor invoice governance, ingestion, reconciliation, attendance attribution and billing evidence.
python-domain-review
Reviews generic Python implementation against task intent, API expectations, simplicity and maintainability.
verified-commit
Creates a task-scoped commit only after validator and exact-tree postflight pass.
testing-patterns
Designs focused, hierarchical, regression, boundary, integration, and end-to-end tests for vertical-slice implementation. Use before and after implementation, and during test review.
test-hierarchy
Classifies and enforces the required unit, integration, local E2E, and cloud E2E test tier for a task or run.
technical-specification
暂无简介