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/mblode/claude-code-search/agents-mdgit clone --depth 1 https://github.com/mblode/claude-code-searchWhat 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 | $0.00658 | $0.00658 |
| Opus 5 | $0.00329 | $0.00329 |
| Sonnet 5 | $0.00132 | $0.00132 |
| Haiku 4.5 | $0.00066 | $0.00066 |
Grade A, and why
claude-code-search 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 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.
How it starts
The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — claude-code-search
ccs, a terminal UI for searching your own coding-agent prompts across past
sessions — fzf or atuin, but for prompt history. Sources: Claude Code, Codex
CLI, Cursor. Published to npm as claude-code-search.
Commands
npm install # setup (requires Node >= 24.11)
npm run dev # tsdown --watch
npm run build # tsdown (CLI + library + dts)
npm run start # node dist/cli.js
npm run lint # ultracite check (oxlint + oxfmt)
npm run check # same as lint
npm run format # ultracite fix
npm run check:types # tsc --noEmit
npm test # vitest run --passWithNoTests
Gates: lint, check:types, test. Smoke: npm run build && npm run start against your own history.
This is a React app that renders to a terminal
The UI is Ink, so src/app.tsx and
src/components/*.tsx are React components whose output is text.
- There is no DOM. Layout comes from Ink's
<Box>; colour fromsrc/utils/color.ts. - Anything written to stdout that is not Ink's render corrupts the frame. Debug to stderr or to a file. CLI data (list/search/json) is the exception: stdout is data, stderr is logs.
- Input is
ink-text-inputplus raw key handling.
Build shape
tsdown.config.ts produces two entries: the binary (dist/cli.js, package.json bin → ccs) with a shebang banner, and the library (dist/index.js + dts). Do not put a shebang in src/cli.ts.
files is dist, example.png, README, LICENSE. Check npm pack --dry-run after touching the build.
Reading session history
Never write to Claude, Codex, or Cursor history files. Parse failures are expected: skip the line, do not throw a stack at the user.
| Source | Layout |
|---|---|
| Claude | ~/.claude/projects/<encoded-cwd>/<uuid>.jsonl (also ~/.config/claude/projects if present). Skip subagents/. |
| Codex | ~/.codex/sessions/**/rollout-*.jsonl and archived_sessions. Skip .jsonl.zst. |
| Cursor | Read-only state.vscdb (composerHeaders + bubbleId type 1). Cwd-scoped unless --source cursor alone. |
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 · 56 lines · 658 tokens per session scan A f7922b81ae9a
claude-code-search AGENTS.md is an instructions file published in the GitHub repository mblode/claude-code-search (5 stars, last pushed 12d ago), licensed MIT. It adds 658 tokens to every session, about $0.0033 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.
Other instructions, from other repositories
cs CLAUDE.md
Instructions for boyter/cs, covering claude.md, project overview, build & test commands, architecture and query pipeline (pkg/search/).
Xapiand AGENTS.md
Instructions for Kronuz/Xapiand, covering agents.md — working in the xapiand codebase, first, the big split: own code vs. vendored, where things live, build & toggles and verifying changes end-to-end (the docs are a test suite).
ketch CLAUDE.md
Instructions for 1broseidon/ketch, covering ketch, architecture, build & test, output format and output flags (all commands).
whoop-mcp CLAUDE.md
Instructions for shashankswe2020-ux/whoop-mcp, covering project: whoop-mcp, tech stack, commands, project structure and code conventions.
moraine AGENTS.md
Instructions for eric-tramel/moraine, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.
huly-mcp CLAUDE.md
Instructions for dearlordylord/huly-mcp, covering project instructions, design principle: llm-first api, project harness (copy to new projects), package manager and verification.