iusztinpaul/squid/squid-clean-docs

Claude Code and agent skill by iusztinpaul/squid

Trust tier 2
  • community
  • ext:[{"type":"hook","description":"Pre/post lifecycle hooks"}]
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Strip redundancy from prose — docs, READMEs, code comments, docstrings — keeping behaviour identical.

Install

npx vskill@latest install iusztinpaul/squid/squid-clean-docs

Semi-extensibleMentions customization, but needs skill-specific knowledge. Read the extensible skills standard

SKILL.md

Clean docs — remove words, keep behaviour

You are the cleaner. You delete redundant prose — markdown docs, code comments, docstrings; you do not touch executable code. Zero behaviour diff is the contract — if a deletion could alter runtime, it is out of scope.

$ARGUMENTS is an optional scope path (docs/, src/ingest/). Empty means the whole repo.

When NOT to use — never touch these without explicit opt-in

ExcludedWhy
docs/adr/, docs/notes/ADRs are immutable historical records; notes are the user's.
CHANGELOG*, LICENSE*Append-only / legally load-bearing.
.github/Issue/PR templates and workflow files are functional surface, not reader prose.
Generated + vendored trees (node_modules/, .venv/, dist/, vendor/, *_pb2.py)Not authored; regenerated on build.
Linter/type directives (# noqa, # type: ignore, # pragma: no cover)These are code, not comments.
Doctest examples (>>> blocks in docstrings)Executable tests, not prose.

Step 1 — Confirm scope

State the default exclusion list above, then ask (one round, AskUserQuestion): what else is off-limits? Do not scan until answered.

Step 2 — Apply the keep/delete test

Read every candidate file whole — never delete on a filename or a skim; a comment can only be judged next to the line it sits on.

A comment or docstring survives only if it states something the code cannot. A constraint, a non-obvious _why_, a limit, a workaround and its issue link. If it narrates _what_ the next line does, it is noise — delete it.

DeleteKeep
# increment the counter above i += 1# Chunk at 200 tokens — the embedding model truncates above 256.
A docstring restating the signature: """Takes a user_id and returns a User.""""""Returns None for soft-deleted users; callers must handle it (see #412)."""
# TODO: refactor this someday (no owner, no issue)# TODO(#88): drop once the v1 endpoint is retired.
# Fixed the bug where... — history belongs in git# Retry 3×: the upstream API 502s under ~5% of cold starts.

Docs: a section dies when it duplicates a section that lives elsewhere. Keep exactly one home for each idea; replace the copy with a one-line cross-reference. A README that repeats CONTRIBUTING.md's install steps loses them and links instead.

Tasks/templates: collapse repeated task files into one template plus the deltas; keep the template.

Step 3 — Plan of attack (the output artifact)

Print one table in chat. Do not write it to disk unless the user asks.

FileWhat goesRule (Step 2)~Lines
src/ingest/chunk.py4 narrating comments, 1 signature docstringrestates code−12
README.mdinstall section duplicating CONTRIBUTING.mdone home per idea−18

Stop and wait for explicit approval. Do not edit before it.

Step 4 — Execute

Apply the approved plan. Commit one category at a time (comments, docs) so any revert is surgical.

Step 5 — Verify

Behaviour must be provably unchanged before hand-off:

  • Test + lint suite green (make pre-commit && make unit-tests, or the project's equivalent) —

docstring deletions can break doctests and doc builds.

  • git diff --stat touches only prose and docstrings — no logic lines.
  • No broken internal links: every path, heading, or anchor you deleted or moved has no remaining

referrers.

Anything red: revert that commit, do not "fix forward".

Step 6 — Hand-off

Report lines removed per category and the verify result. If the README was reorganised, say what moved where.

When run unattended (a scheduled routine), also send that report with PushNotification — nobody reads the session.

Notes on shape

  • Removing beats rewriting. Prefer deleting a paragraph to shortening it.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Extension points

  • hookPre/post lifecycle hooks

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/iusztinpaul/squid/squid-clean-docs/badge)](https://verified-skill.com/skills/iusztinpaul/squid/squid-clean-docs)

Something wrong with this skill or its page? Tell us.