authoring-instructions
/authoring-instructions
Tworzy i audytuje pliki instrukcji czytane przez agentów: reguły, główny AGENTS.md z cienkim mostkiem CLAUDE.md oraz hooki zdarzeń (hooks.json i skrypty).
- Autor
- richfrem ↗
- Repozytorium
- agent-plugins-skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 24.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 23
- Skrypty
- 8
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 24.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
- 8 skryptów w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.
Skrypty: scripts/audit_agents_md.py, scripts/audit_rule.py, scripts/hook_linter.py, scripts/scaffold_hook.py, scripts/test_hook.py, scripts/validate_hook_schema.py, scripts/validate_write.py, tests/test_hooks.py
Co robi i kiedy
Tworzy i audytuje pliki instrukcji czytane przez agentów: reguły, główny AGENTS.md z cienkim mostkiem CLAUDE.md oraz hooki zdarzeń (hooks.json i skrypty).
Argumenty
/authoring-instructions [rule|agents-md|hook] [create|audit] [path]
Oryginalny opis (EN)
Creates and audits the instruction files agents read: agent rules (rules/*.md), the root AGENTS.md with its thin CLAUDE.md bridge, and event-driven hooks (hooks.json plus handler scripts). Use when the user wants to write, check, trim or migrate a rule, AGENTS.md or CLAUDE.md, or to add or lint a lifecycle hook such as PreToolUse or SessionStart. Not for skills (use authoring-skills), whole plugins (use authoring-plugins) or GitHub workflows (use authoring-github-workflows).
Jak działa
- Contents
- Constraints
- Modes
- Quick start
- Workflow
- Verification
- References
Podgląd SKILL.md
---
name: authoring-instructions
plugin: agent-extension-authoring
description: >-
Creates and audits the instruction files agents read: agent rules (rules/*.md), the root AGENTS.md with its thin CLAUDE.md bridge, and event-driven hooks (hooks.json plus handler scripts). Use when the user wants to write, check, trim or migrate a rule, AGENTS.md or CLAUDE.md, or to add or lint a lifecycle hook such as PreToolUse or SessionStart. Not for skills (use authoring-skills), whole plugins (use authoring-plugins) or GitHub workflows (use authoring-github-workflows).
argument-hint: "[rule|agents-md|hook] [create|audit] [path]"
allowed-tools: Bash, Read, Write
---
# Authoring Instructions (`authoring-instructions`)
Writes lean, invariant-driven instruction files and audits them deterministically: rules, root `AGENTS.md` / `CLAUDE.md`, and plugin hooks.
## Contents
- [Constraints](#constraints) • [Modes](#modes) • [Quick start](#quick-start) • [Workflow](#workflow) • [Verification](#verification) • [References](#references)
## Constraints
1. **Rules are constraints, skills are procedures.** A rule states MUST / NEVER / ALWAYS bounds in 30–70 lines (ceiling 80); anything with ordered steps becomes a skill.
2. **Timeless and portable**: no calendar dates, commit SHAs, usernames or absolute machine paths in a rule or root file.
3. **Never overwrite** an existing `AGENTS.md` or `CLAUDE.md`; propose a diff. `CLAUDE.md` stays a bridge of 5 lines or fewer.
4. **Audits are read-only.** They report findings; changes are reviewed edits.
5. **Hooks** use `python3 ... || python ...` commands anchored on `${CLAUDE_PLUGIN_ROOT}`, exit silently when the repository is not initialised, and stay fast.
6. APM does not deploy a plugin's `rules/` folder (instructions deploy from `.apm/instructions/`), so rules need no link or registration step.
## Modes
| Mode | Use when | Entry point | Reference |
|---|---|---|---|
| `create-rule` | New rule file | write the canonical structure | [create-rule.md](references/modes/create-rule.md) |
| `audit-rule` | Check rules | `scripts/audit_rule.py` | [audit-rule.md](references/modes/audit-rule.md) |
| `create-agents-md` | Scaffold or migrate root files | `assets/templates/AGENTS.md.jinja` | [create-agents-md.md](references/modes/create-agents-md.md) |
| `audit-agents-md` | Check size, timelessness, disclosure | `scripts/audit_agents_md.py` | [audit-agents-md.md](references/modes/audit-agents-md.md) |
| `hook` | Add, lint or test a lifecycle hook | `scripts/scaffold_hook.py` | [create-hook.md](references/modes/create-hook.md) |
## Quick start
```bash
python3 scripts/audit_rule.py <rules-dir-or-file> --json
python3 scripts/audit_agents_md.py AGENTS.md --strict
python3 scripts/scaffold_hook.py --event PreToolUse --name check-command --plugin-root plugins/<plugin>
```
## Workflow
1. Pick the mode and read its reference before running anything.
2. Create modes: draft from the canonical structure or template, then run the matching audit.
3. Hook mode: scaffold, implement the handler, then run `scripts/validate_hook_schema.py` and `scripts/hook_linter.py`; exercise it with `scripts/test_hook.py`.
4. On any failure follow that mode's section in [fallback-tree.md](references/fallback-tree.md).
## Verification
```bash
python3 scripts/audit_rule.py <rule.md> --strict
python3 scripts/audit_agents_md.py . --all --strict
python3 scripts/validate_hook_schema.py plugins/<plugin>/hooks/hooks.json
```
Tests: `python3 -m pytest tests`. Check the mode's section in [acceptance-criteria.md](references/acceptance-criteria.md).
## References
- Modes: [create-rule](references/modes/create-rule.md) • [audit-rule](references/modes/audit-rule.md) • [create-agents-md](references/modes/create-agents-md.md) • [audit-agents-md](references/modes/audit-agents-md.md) • [hook](references/modes/create-hook.md)
- Detail: [agents-md-contract.md](references/agents-md-contract.md) • [hooks-spec.md](references/hooks-spec.md) • [hook-patterns.md](references/hook-patterns.md) • [hook-advanced.md](references/hook-advanced.md) • [hook-migration.md](references/hook-migration.md)
- Gates: [acceptance-criteria.md](references/acceptance-criteria.md) • [fallback-tree.md](references/fallback-tree.md)
- Template: [AGENTS.md.jinja](assets/templates/AGENTS.md.jinja)
Pliki w paczce (23)
- SKILL.md 4 KB
- assets/templates/AGENTS.md.jinja 1 KB
- evals/evals.json 3 KB
- references/acceptance-criteria.md 2 KB
- references/agents-md-contract.md 4 KB
- references/fallback-tree.md 2 KB
- references/hook-advanced.md 11 KB
- references/hook-migration.md 9 KB
- references/hook-patterns.md 8 KB
- references/hooks-spec.md 4 KB
- references/modes/audit-agents-md.md 2 KB
- references/modes/audit-rule.md 2 KB
- references/modes/create-agents-md.md 3 KB
- references/modes/create-hook.md 3 KB
- references/modes/create-rule.md 3 KB
- scripts/audit_agents_md.py 15 KB
- scripts/audit_rule.py 12 KB
- scripts/hook_linter.py 7 KB
- scripts/scaffold_hook.py 4 KB
- scripts/test_hook.py 10 KB
- scripts/validate_hook_schema.py 9 KB
- scripts/validate_write.py 3 KB
- tests/test_hooks.py 3 KB
Wersje
W bazie nie ma innych skilli o nazwie authoring-instructions.
Migawka bazy z 28.10.2026 · wersja paczki c2a2477dfeb2 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100authoring-skills
Tworzy, audytuje i poprawia skille agentów (foldery SKILL.md), dodaje środowisko Docker, integrację z serwerem MCP i pozwala poznać specyfikację Agent Skills z katalogiem 64 wzorców. - 100rc-writing-for-agents
Pisze treści czytane przez agentów, czyli skille, AGENTS.md i hooki, oraz audytuje koszt kontekstu przy ich tworzeniu lub skracaniu. - 100authoring-plugins
Tworzy szkielety i audytuje całe pluginy agentów: katalog, manifest, README, folder skilli, a także weryfikuje strukturę, plugin.json, komponenty i przenośność istniejących pluginów. - 100plugin-authoring
Pomaga tworzyć, modyfikować i debugować wtyczki Claude Code: schematy, szablony, walidacja i rozwiązywanie problemów dla plugin.json, marketplace.json oraz katalogów commands, agents, skills i hooks. - 100plugin-authoring
Pomaga tworzyć, modyfikować i debugować wtyczki Claude Code: schematy, szablony, walidacja i rozwiązywanie problemów dla plugin.json, marketplace.json oraz katalogów commands, agents, skills i hooks. - 100plugin-authoring
Tworzy wtyczki Claude Code pakujące skille, agentów, hooki i serwery MCP w dystrybuowalne paczki: manifest plugin.json, struktura katalogów, przestrzenie nazw i dystrybucja.