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 agents/invariance-ai/gps/claudegit clone --depth 1 https://github.com/invariance-ai/gpsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/invariance-ai/gps/claude)<a href="https://agentmods.dev/agents/invariance-ai/gps/claude"><img src="https://agentmods.dev/badge/agents/invariance-ai/gps/claude.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.01122 |
| Opus 5 | $0.00000 | $0.00561 |
| Sonnet 5 | $0.00000 | $0.00224 |
| Haiku 4.5 | $0.00000 | $0.00112 |
Grade A, and why
claude 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 4d 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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claude Code
GPS installs as a normal Claude Code project integration: instructions, a skill, hooks, and an MCP server. The point is not to make you memorize new commands. The point is to teach Claude to ask GPS for repo memory before meaningful edits, and to capture useful lessons after the work.
npx -y @invariance/gps setup --yes --with-claude
Writes:
CLAUDE.md— appends a<!-- gps:start -->...<!-- gps:end -->block with the GPS workflow and standing preference guidance. Idempotent: re-runs replace only the managed block..claude/skills/gps/SKILL.md— the GPS skill. Claude Code auto-loads it and learns when to call GPS..claude/settings.json— five non-blocking hooks. See below..mcp.json— registers thegpsMCP server (gps serve). Merged with any existingmcpServersentries. This is what makes themcp__gps__prepare_edittool callable.
If you want the portable piece, copy only .claude/skills/gps/SKILL.md. That is the Claude Code skill that teaches the loop. The full setup command also writes CLAUDE.md, hooks, and MCP so Claude does not just know the rules, it gets context injected and memory captured automatically.
What Claude learns
The installed skill and CLAUDE.md block tell Claude to use GPS like a repo-memory layer:
gps prepare --intent "<what I am about to change>"
gps prepare <symbol> --intent "<short intent>"
gps brief
gps remember "<hard-won repo fact>"
gps lessons record "<one sentence>"
When MCP is available, Claude can call mcp__gps__prepare_edit instead of shelling out. The CLI and MCP tool return the same kind of brief: relevant files, callers, likely tests, invariants, notes, decisions, and preferences.
Use the CLI when it is simpler. Use MCP when Claude already has the tool exposed. Do not call both for the same brief.
The five hooks
| Hook | When | What it does |
|---|---|---|
SessionStart (startup|resume) |
Session opens | gps index (rebuild graph), gps feature clear-active, gps session start, print gps preferences --markdown |
UserPromptSubmit |
Every prompt | gps capture-preference --emit (catches "from now on…" rules), gps context-from-prompt (auto-inject context for named symbols) |
PreToolUse Edit|MultiEdit|Write |
Before any edit | gps index — keep graph fresh |
PostToolUse Bash |
After every Bash | If exit code ≠ 0, gps record-failure --kind bash against the last-prepared symbol |
Stop |
Turn end | gps attach --hook-stdin (distill the session transcript into Decisions/Questions — auto-persists with an API key, else stashes a prompt under .gps/pending-distill/), gps feature attribute --git-diff (only if gps validate passes — stale graphs poison attribution), gps session end |
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.
- 4d ago First seen · 82 lines · 0 tokens per session scan A 1c19ea06dbdc
claude is an agent published in the GitHub repository invariance-ai/gps (19 stars, last pushed 3mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,122 tokens. 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 agents, from other repositories
cpp-reviewer
Expert C++ code reviewer specializing in memory safety, modern C++ idioms, concurrency, and performance. Use for all C++ code changes. MUST BE USED for C++ projects.
context
You are the Context agent. Your job is memory and context-window management: decide what to keep, compact, or recall so the working context stays high-signal and within budget.
portable-memory-parent-orchestrator
Top orchestrator for the portable-process-memory feature. Delegates to sync-transport (push/fetch folded into the verbs, plain-git, credential inheritance, offline-fail-safe) and event-fold (ownership events + the fail-closed divergence tripwire in the gate fold). Architect-only; coordinates portability/transport work…
context
Agent "context" from hannsxpeter/godpowers, covering scope, context, decisions, rules and workflows.
neon-drizzle-expert
Expert in Neon PostgreSQL (v17), Drizzle ORM (v0.44.4), and Zod (v4.0.17) schema validation for production memory systems. Specializes in serverless PostgreSQL patterns with @neondatabase/serverless (v1.0.1), type-safe database operations, and migration strategies.
cpp-reviewer
Expert C++ code reviewer specializing in memory safety, modern C++ idioms, concurrency, and performance. Use for all C++ code changes. MUST BE USED for C++ projects.