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.
npx agentmods add instructions/mareurs/codescout/claude-mdgit clone --depth 1 https://github.com/mareurs/codescoutWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.10376 | $0.10376 |
| Opus 5 | $0.05188 | $0.05188 |
| Sonnet 5 | $0.02075 | $0.02075 |
| Haiku 4.5 | $0.01038 | $0.01038 |
Grade B, and why
codescout CLAUDE.md scanned grade B with 1 finding 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
- Auto-detects codescout from `.mcp.json` or `~/.claude/settings.json` How it starts
The opening of the file, as written. The whole thing — 678 lines — stays where its author put it; the contents beside it link to each section on GitHub.
codescout
Rust MCP server giving LLMs IDE-grade code intelligence — symbol-level navigation, semantic search, git integration. Inspired by Serena.
You are a proficient Rust developer. You follow all known good/scalable patterns. You are honest and recognize your limits and your mistakes, you own them. If you are not sure, you always ask me for feedback.
Development Commands
See codescout memory development-commands for the full command reference.
Always run cargo fmt, cargo clippy, and cargo test before completing any task.
On our current stack the release build is cargo rb, not cargo build --release (the latter is the lean default repo cloners get). See the development-commands memory for what the alias compiles in and why.
To test changes via the live MCP server, run cargo rb, then restart the server with /mcp. The MCP server launches ~/.cargo/bin/codescout (see the command in ~/.claude/.claude.json), which is a symlink → target/release/codescout, so a release build alone updates the live binary — no cargo install needed. Dev builds (cargo test/cargo build) are a separate artifact and are NOT picked up.
If the symlink is missing (e.g. after cargo clean removed target/, or a fresh checkout), recreate it once:
ln -sf "$(pwd)/target/release/codescout" ~/.cargo/bin/codescout
Without the symlink, ~/.cargo/bin/codescout is a stale installed copy and /mcp reconnects keep loading old code even after a successful build. (Symlink, not hardlink: cargo build rename-replaces the file, so a hardlink would resolve to the old inode.)
Bug Tracking
Per-file bug tracking lives in docs/issues/. Every bug noticed during work gets its own file, copied from docs/issues/_TEMPLATE.md.
- Path:
docs/issues/YYYY-MM-DD-<slug>.mdfor active bugs;docs/issues/archive/only after the fix has shipped tomaster(verify viagit branch --contains <fix-sha>). - Slug: short kebab-case noun-phrase (3–6 words), e.g.
edit-code-insert-mid-function,reindex-cascade-delete-data-loss. - Status field in frontmatter:
open|investigating|fixed|mitigated|wontfix|zombie(semantics in_TEMPLATE.md's header comment).zombie= no longer observed, root cause unconfirmed; pair withlast_observed:and a re-open trigger. closed:date in frontmatter alongside any offixed/mitigated/wontfix.
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.
- 2d ago First seen · 678 lines · 10,376 tokens per session scan B f0db7bb8b4f9
codescout CLAUDE.md is an instructions file published in the GitHub repository mareurs/codescout (21 stars, last pushed 2d ago), licensed MIT. It adds 10,376 tokens to every session, about $0.0519 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
roam-code AGENTS.md
Instructions for Cranot/roam-code, covering agents.md — roam-code development guide, what this project is, documentation hub, where files go (private vs public) and quality discipline (from internal/dogfood/ + agi-in-md).
Ctxo CLAUDE.md
Instructions for alperhankendi/Ctxo, covering claude.md — ctxo, project overview, quick reference, dev (pnpm workspace) and usage (consumer).
agent-lsp CLAUDE.md
Instructions for blackwell-systems/agent-lsp, a project described as: MCP server that orchestrates language servers into agent-native workflows. 65 tools, 30 CI-verified languages.
jcodemunch-mcp CLAUDE.md
Claude Code instructions for jgravelle/jcodemunch-mcp, covering jcodemunch-mcp — project brief, current state, key files, cli subcommands and architecture notes.
leantoken AGENTS.md
Instructions for morluto/leantoken, covering repository guidance, architecture, development, change-specific validation and contributions.
jcodemunch-mcp AGENTS.md
AGENTS.md instructions for jgravelle/jcodemunch-mcp, covering code exploration policy, session-aware routing and model-driven tool tiering.