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/cmg8431/claude-dejavu/agents-mdgit clone --depth 1 https://github.com/cmg8431/claude-dejavuWhat 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 | $0.00589 | $0.00589 |
| Opus 5 | $0.00295 | $0.00295 |
| Sonnet 5 | $0.00118 | $0.00118 |
| Haiku 4.5 | $0.00059 | $0.00059 |
Grade A, and why
claude-dejavu AGENTS.md scanned grade A 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 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
- **TypeScript launcher** (`packages/cli/`) — platform detection + `spawnSync` execution How it starts
The opening of the file, as written. The whole thing — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Agent Guidelines
Commit Message Convention
<type>(<scope>): <description>
Types
| Type | Description |
|---|---|
feat |
New feature |
fix |
Bug fix |
refactor |
Code refactoring (no behavior change) |
docs |
Documentation only |
test |
Adding or updating tests |
chore |
Maintenance tasks |
perf |
Performance improvements |
Architecture Overview
claude-dejavu is a Rust + TypeScript hybrid monorepo (same pattern as codagotchi):
- Rust library (
crates/dejavu-core/) — JSONL parser, detectors, DB, rule engine - Rust binary (
crates/dejavu-cli/) — CLI with subcommands (init, scan, list, stats) - TypeScript launcher (
packages/cli/) — platform detection +spawnSyncexecution
Core Loop
Session logs (.jsonl) → Parser → Detectors → Patterns (SQLite) → Rules → CLAUDE.md patch
│
Rule effectiveness scoring ◀──────────────────────┘
Detectors (v0)
- Revert Cycle — Edit → revert → re-edit on same file
- Repeated Error — Same error appears across sessions, same fix each time
- Silent Fix — User silently corrects Claude's output without asking
Key Design Decisions
- SQLite with WAL mode.
~/.local/share/claude-dejavu/dejavu.db - CLAUDE.md metadata in HTML comments. Claude ignores them, dejavu reads/writes them.
- No traits for detectors. Each detector is a standalone
detect(&[ParsedSession])function. anyhow::Resultfor all fallible functions.
Adding a New Detector
- Create
detector/{name}.rswithpub fn detect(sessions: &[ParsedSession]) -> Vec<Detection> - Add
pub mod {name};todetector/mod.rs - Add call in
run_all_detectors()
Adding a New CLI Command
- Create
commands/{name}.rswithpub fn run(...) -> Result<()> - Add
pub mod {name};tocommands/mod.rs - Add variant to
Commandsenum inmain.rs - Add match arm in
main()
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 · 69 lines · 589 tokens per session scan A af61766f559f
claude-dejavu AGENTS.md is an instructions file published in the GitHub repository cmg8431/claude-dejavu (2 stars, last pushed 3mo ago), licensed MIT. It adds 589 tokens to every session, about $0.0029 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
claude-mem CLAUDE.md
Instructions for thedotmack/claude-mem, covering claude-mem: ai development instructions, build, file locations, requirements and documentation.
claude-mem copilot-instructions.md
Instructions for thedotmack/claude-mem, a project described as: Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot…
vexic AGENTS.md
Instructions for PyroDonkey/vexic, covering vexic agent instructions, project, architecture boundaries, package boundary and host ports.
longhand CLAUDE.md
Instructions for Wynelson94/longhand, covering longhand mcp tools — efficient usage guide, quick decision tree, anti-patterns (avoid), tool pairing patterns and pattern b: recall past work across sessions (1-2 calls).
claude-mem-sync CLAUDE.md
Instructions for lopadova/claude-mem-sync, covering claude.md — claude-mem-sync, what this project is, architecture, project structure and commands.
junto-memory CLAUDE.md
Instructions for tlemmons/junto-memory, covering junto memory (mcp server), claude identity (required — do this first), project roster (post-cutover 2026-05-07), rename aliases (active 30 days from 2026-05-07) and key files.