project-standards-practice
Read AGENTS.md/SOP, consolidate standards, apply to tasks.
- License 许可
- MIT
- 所属仓库
- zapabob/Skills
- 最近更新
- 2026-10-26
- Stars
- ★ 2
- Forks
- 0
01 前言 Front Matter
- name
- project-standards-practice
- author
- Hermes Agent
- license
- MIT
- version
- 0.1.0
- metadata
- { "hermes": { "tags": [ "coding-standards", "AGENTS.md", "SOP", "best-practices", "workflow" ], "related_skills": [ "hermes-agent-skill-authoring", "systematic-debugging", "cross-platform-path" ] } }
- platforms
- [ "windows", "linux", "macos" ]
- description
- Read AGENTS.md/SOP, consolidate standards, apply to tasks.
03 SKILL.md 正文
Project Standards Practice
Read the project's AGENTS.md and SOP documents, distill the key coding standards, and apply them to subsequent tasks.
When to Use
- Starting a new coding task in an unfamiliar codebase
- User asks to "read the AGENTS.md and SOP" or "check the project standards"
- Beginning work on a repository with established conventions
- Before making changes that need to respect project-wide invariants
Prerequisites
None. The skill operates on locally available documentation.
How to Run
-
Locate the standards documents:
- Primary:
AGENTS.mdat repository root (usually the main SOP) - Secondary: search for
SOP*,*.sop, orSTANDARD*files - Check
_docs/directory for implementation records
- Primary:
-
Read and inventory:
- Use
read_fileon AGENTS.md (may be large — read in chunks if needed) - Use
search_filesto find specific topics (e.g., "Windows", "testing", "profiles") - Note section headings to understand document structure
- Use
-
Distill into actionable summary:
- Extract concrete rules (not vague guidance)
- Group by theme: paths, testing, platform quirks, skill authoring, etc.
- Flag anything that directly constrains how you'll work
-
Apply to current task:
- Reference the consolidated standards when making decisions
- Flag any conflicts between standards and the current task
Quick Reference
| Document | Typical Location | Purpose |
|---|---|---|
| AGENTS.md | repo root | Master guide & assistant mirrors |
| CLAUDE.md / GEMINI.md | repo root | Mirror guides for other assistants |
| _docs/ | repo root | Implementation audit logs |
| skills/ | repo root | Built-in skill standards |
Procedure
Step 1: Find the standards
# Search for SOP-like documents
search_files(pattern="AGENTS.md", target="files")
search_files(pattern="*SOP*", target="files")
search_files(pattern="*.sop", target="files")
Step 2: Read the primary document
If AGENTS.md exists at root, read it. For large files, read in sections:
read_file(path="AGENTS.md", limit=500)
# If truncated, continue with offset=501
Step 3: Search for specific topics
Focus on sections relevant to your task:
# Windows-specific pitfalls
search_files(pattern="Windows|os.kill|UTF-8|pathlib", path=".", file_glob="*.md")
# Testing standards
search_files(pattern="test|pytest|mock", path=".", file_glob="*.md")
# Profile safety
search_files(pattern="get_hermes_home|profile", path=".", file_glob="*.md")
Step 4: Consolidate
Produce a concise summary organized by theme. Include:
- Hard requirements (must follow)
- Strong conventions (should follow)
- Platform-specific notes (Windows/Linux/macOS)
- Anti-patterns (what NOT to do)
Step 5: Apply
Reference the consolidated standards throughout the current task. When in doubt, check the original document.
Pitfalls
- Only reading the first part of AGENTS.md — it's often long (1000+ lines). Read the whole thing or search for relevant sections.
- Treating guidelines as optional — AGENTS.md often contains hard invariants (e.g., "Sacred Invariants" section).
- Missing the "What We Reject" section — equally important as "What We Want"; tells you what NOT to do.
- Not checking for platform-specific sections — Windows pitfalls are often called out separately.
- Forgetting skill authoring standards — if creating skills, the HARDLINE section is mandatory.
Verification
- [ ] Read AGENTS.md (or equivalent SOP document) in full or via targeted search
- [ ] Searched for platform-specific sections (Windows/Linux/macOS)
- [ ] Identified hard requirements vs. conventions
- [ ] Consolidated standards into a usable reference
- [ ] Applied standards to current task decisions
Related Skills
hermes-agent-skill-authoring— if creating/editing skillssystematic-debugging— for root cause investigationcross-platform-path— for path handling across OSes
04 同仓库更多 Skills
xurl
X/Twitter via xurl CLI: post, search, DM, media, v2 API.
writing-plans
Write implementation plans: bite-sized tasks, paths, code.
whisper
OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual ASR.
weights-and-biases
W&B: log ML experiments, sweeps, model registry, dashboards.
watchers
Poll RSS, JSON APIs, and GitHub with watermark dedup.
serving-llms-vllm
vLLM: high-throughput LLM serving, OpenAI API, quantization.