claude-ruby-grape-rails plugin-review.instructions.md

A set of rules for reviewing files in coding-agent plugins. It covers agent instructions, skills, and hooks, which are files that shape how an agent works.

In plain words
What is it for?
Use it to review plugin documentation and configuration, check YAML settings for agent definitions, and suggest clearer imperative wording.
Why use it?
It helps catch instructions that agents may ignore or interpret poorly, such as narrative explanations, unsupported settings, and misplaced command comments.

Instructions file for GitHub Copilot

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/slbug/claude-ruby-grape-rails/plugin-review
Clone the repo
git clone --depth 1 https://github.com/slbug/claude-ruby-grape-rails

Made for: GitHub Copilot.

Per session 5,496 This file is loaded in full into every session.
When invoked 5,496 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.05496 $0.05496
Opus 5 $0.02748 $0.02748
Sonnet 5 $0.01099 $0.01099
Haiku 4.5 $0.00550 $0.00550

Measured 2d ago against content hash 70e1a05c4104, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

claude-ruby-grape-rails plugin-review.instructions.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.github/instructions/plugin-review.instructions.md · 406 lines

How it starts

The opening of the file, as written. The whole thing — 406 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Plugin File Review Rules

Audience: Agents, Not Humans

Plugin docs (agents/, skills/, references/) load into agent context at runtime. Flag prose that violates the imperative-only rule:

  • Tutorial narration ("first do X, then Y") → flag, suggest imperative rewrite
  • Reasoning preludes before commands → flag
  • # thinking/checklist lines inside Bash command bodies (preference #6) → flag, suggest markdown table or prose lead-in
  • Long explanatory paragraphs where a table fits → flag

Exception files (human-facing, narrative OK): README.md, CHANGELOG.md, scripts/, lab/eval/.

Agent Conventions (plugins/**/agents/*.md)

  • Agents use YAML frontmatter. Common: name, description, model, disallowedTools, omitClaudeMd, skills, memory. Also valid per CC docs: tools, effort, maxTurns, background, isolation. The general subagent reference (https://docs.claude.com/en/docs/claude-code/sub-agents) also lists color and initialPrompt, but the plugin-supported set documented at https://docs.claude.com/en/docs/claude-code/plugins-reference does NOT include them — CC silently drops these on plugin-shipped agents. Do NOT add them to plugin agents under plugins/**/agents/
  • Prefer denylist-only (disallowedTools:) over allowlist (tools:). A missing tools: field is intentional — agents inherit all tools minus those in disallowedTools
  • Artifact-writing agents: disallowedTools: Edit, NotebookEdit, Agent, EnterWorktree, ExitWorktree, Skill
  • Conversation-only agents add Write to the above
  • NO agent declares or invokes Agent — orchestration lives in skill bodies (main session). See .claude/rules/agent-development.md § "Subagents Are Leaf Workers" for full doctrine.
  • Agents with intentionally narrow tool sets (web-researcher, output-verifier, ruby-gem-researcher) use tools: allowlists — this is correct
  • omitClaudeMd: true is correct for specialist agents that do not need contributor CLAUDE.md context
  • Do NOT flag permissionMode as missing — Claude Code ignores it on plugin agents
  • Model tiers: opus for security-critical agents, sonnet for most specialists, haiku for mechanical tasks. opusplan is a session-level /model alias only (the subagent frontmatter table at https://docs.claude.com/en/docs/claude-code/sub-agents does NOT list it as a valid subagent model value) — do NOT flag its absence from agent frontmatter as a gap
  • Descriptions must be <= 250 characters

Read the full file on GitHub · 406 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 406 lines · 5,496 tokens per session scan A 70e1a05c4104

Subscribe to this mod's changes

claude-ruby-grape-rails plugin-review.instructions.md is an instructions file published in the GitHub repository slbug/claude-ruby-grape-rails (7 stars, last pushed 3d ago), licensed MIT. It adds 5,496 tokens to every session, about $0.0275 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other instructions, from other repositories