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/justrach/codedb/agents-mdgit clone --depth 1 https://github.com/justrach/codedbWhat 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 | $0.00509 | $0.00509 |
| Opus 5 | $0.00254 | $0.00254 |
| Sonnet 5 | $0.00102 | $0.00102 |
| Haiku 4.5 | $0.00051 | $0.00051 |
Grade A, and why
codedb AGENTS.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 yesterday.
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.
How it starts
The opening of the file, as written. The whole thing — 47 lines — stays where its author put it; the contents beside it link to each section on GitHub.
codedb Agent Guidelines
What codedb is (and isn't)
codedb is a code-intelligence and context tool, not an editor. It exists to
help agents find and understand code — structural search, symbol/caller
lookup, dependency graph, outlines, and task-shaped context — so they can edit
with their own native tools. codedb has no edit capability: the old
codedb_edit fallback (and the HTTP POST /edit endpoint, the post-edit
linter, and codedb_diagnostics) was removed because MCP edit calls bypass the
client's own undo/rewind tracking. Keep that framing consistent across tool
descriptions, the MCP initialize instructions, the README, and these docs.
Review guidelines
- Flag any security issues: injection, file traversal, untrusted input, secret exposure
- Verify that sensitive files (.env, .pem, .key, credentials) are excluded from indexing AND search
- Check that telemetry behavior matches documentation claims
- Flag any regression in benchmark-critical paths (threshold: 10%)
- Treat P1 issues as merge-blocking
- Verify new language parsers handle malformed input gracefully (braces in strings, unterminated comments)
- Check that installer scripts don't execute untrusted code or skip verification
Pre-merge verification
Run these before merging any MCP-related change:
zig build test # unit tests
python3 scripts/e2e_mcp_test.py \
--binary zig-out/bin/codedb \
--project /path/to/codedb # E2E MCP scenarios
e2e_mcp_test.py covers three scenarios:
- issue-346 regression — spawn from cwd=
/, roots handshake, tools return real data - Normal mode — explicit positional root (
codedb <path> mcp), immediate scan - No-roots client — spawn from
/with no roots capability, stays alive gracefully
Security-sensitive areas
src/watcher.zig— file indexing skip lists (secrets must be excluded)src/mcp.zig— file read/search (path traversal, scope boundaries)src/telemetry.zig— data collection and transmission (must match docs)src/snapshot.zig— sensitive file filteringinstall/install.sh— binary download and config modification
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.
- yesterday First seen · 47 lines · 509 tokens per session scan A c7eeff4746ce
codedb AGENTS.md is an instructions file published in the GitHub repository justrach/codedb (1,371 stars, last pushed 2d ago), licensed BSD-3-Clause. It adds 509 tokens to every session, about $0.0025 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 2026-08-30.
Other instructions, from other repositories
capsule CLAUDE.md
Instructions for endorhq/capsule, covering claude.md, project overview, tech stack, commands and web app (packages/web).
gmail-mcp GEMINI.md
Gemini CLI instructions for hahaha-saygex/gmail-mcp, covering gmail mcp guidelines for gemini, available operations, email operations (9), label management (5) and filter management (5).
gmail-mcp CLAUDE.md
Claude Code instructions for hahaha-saygex/gmail-mcp, covering gmail mcp guidelines for claude, available operations, email operations (10), label management (5) and filter management (5).
hindsight CLAUDE.md
Instructions for vectorize-io/hindsight, covering claude.md, project overview, development commands, local development (api + ui) and start both api server and control plane ui.
claude-code-settings copilot-instructions.md
Instructions for feiskyer/claude-code-settings, covering claude.md, environment setup, required dependencies, configuration and skills.
goodboy AGENTS.md
Instructions for akhayam99/goodboy, covering agents.md, where new code goes, naming, components and exports and store selectors and memoization.