agentmods.dev

pi-extensions CLAUDE.md

Repository instructions for coding agents working on a TypeScript npm workspaces monorepo. They describe commands, package layout, architecture, and data flow for the pi-extensions project.

In plain words
What is it for?
Use them when changing pi-continuous-learning or other packages, running project checks, understanding extension hooks and commands, or following the repository's build and CI workflow.
Why use it?
They tell the agent how to test, lint, type-check, build, and navigate a project containing several related packages.

Instructions file

130 9d ago A scan 870 tokens original MIT
View the source file MattDevy/pi-extensions
Install

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.

agentmods
npx agentmods add instructions/mattdevy/pi-extensions/claude-md
Clone the repo
git clone --depth 1 https://github.com/MattDevy/pi-extensions
Per session 870 This file is loaded in full into every session.
When invoked 870 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00870 $0.00870
Opus 5 $0.00435 $0.00435
Sonnet 5 $0.00174 $0.00174
Haiku 4.5 $0.00087 $0.00087

Measured 2d ago against content hash 9ca1bddc8bc5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

pi-extensions CLAUDE.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.

CLAUDE.md · 70 lines

How it starts

The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Commands

npm test                          # run all package tests (workspaces)
npm test -w packages/pi-continuous-learning -- src/foo.test.ts  # single file
npm test -w packages/pi-continuous-learning -- -t "pattern"     # by name pattern
npm run typecheck                 # type-check all packages
npm run lint                      # ESLint on all packages
npm run check                     # tests + lint + typecheck (mirrors CI)
npm run lint:mega                 # run MegaLinter locally (requires Docker)
npm run lint:mega:fix             # run MegaLinter and auto-fix formatting
npm run build                     # compile all packages to dist/

Architecture

This is an npm workspaces monorepo. The pi-continuous-learning package lives under packages/pi-continuous-learning/. Its entry point (packages/pi-continuous-learning/src/index.ts) exports a default function that receives ExtensionAPI and registers hooks and commands.

Data flow

Pi session (extension)               Background analyzer (pi-cl-analyze CLI)
──────────────────────               ──────────────────────────────────────
Hooks observe session events    →    Reads observations.jsonl per project
  writes to observations.jsonl       Calls Haiku LLM to find patterns
                                     Creates/updates instinct .md files
Before next agent start         ←
  high-confidence instincts injected into system prompt
  feedback loop records which instincts were active
  confidence adjusted by real outcomes

The analyzer runs as a separate background process (cron/launchd), never inside a Pi session.

Key modules

All source lives under packages/pi-continuous-learning/src/:

  • Observers (tool-observer.ts, session-observer.ts, prompt-observer.ts) — capture session events and write observations.jsonl
  • Instinct store (instinct-store.ts, instinct-parser.ts, instinct-loader.ts) — CRUD for markdown instinct files (YAML frontmatter + body)
  • Injector (instinct-injector.ts, active-instincts.ts) — selects high-confidence instincts and injects them into the system prompt before each agent start
  • Confidence (confidence.ts, instinct-decay.ts) — scoring and TTL-based decay
  • CLI analyzer (cli/analyze.ts) — standalone background process with lockfile guard, 5-minute global timeout, structured JSON logging
  • Commands (src/commands/) — slash commands registered with Pi
  • Tools (instinct-tools.ts) — LLM-callable tools for instinct CRUD
  • Prompts (prompts/) — system and user prompts for the LLM analyzer, consolidation, and evolution passes

Read the full file on GitHub · 70 lines

Changes

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.

  1. 2d ago First seen · 70 lines · 870 tokens per session scan A 9ca1bddc8bc5

Subscribe to this mod's changes

pi-extensions CLAUDE.md is an instructions file published in the GitHub repository MattDevy/pi-extensions (130 stars, last pushed 9d ago), licensed MIT. It adds 870 tokens to every session, about $0.0043 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.

LLMaipi-coding-agentpi-extensionpi-extensions
Related

Other instructions, from other repositories

pi-extensions AGENTS.md

Instructions for narumiruna/pi-extensions, covering repository guidelines, documentation and communication, repository structure, commands and tooling and dependency safety.

narumiruna/pi-extensions · 3,342 tokens

kendex AGENTS.md

AGENTS.md instructions for vanillagreencom/kendex, covering kendex and code review rules.

vanillagreencom/kendex · 1,050 tokens

kendex copilot-instructions.md

Copilot instructions for vanillagreencom/kendex, covering kendex, code review calibration, what to raise, what not to raise and reply contract (context for reading threads).

vanillagreencom/kendex · 459 tokens

kendex skills-and-agents.instructions.md

Instructions for vanillagreencom/kendex, a project described as: Package manager for agents, skills, hooks, and extensions. Author once, install on every harness. QOL features included.

vanillagreencom/kendex · 259 tokens

kendex tests.instructions.md

Instructions for vanillagreencom/kendex, a project described as: Package manager for agents, skills, hooks, and extensions. Author once, install on every harness. QOL features included.

vanillagreencom/kendex · 174 tokens

kendex crates.instructions.md

Instructions for vanillagreencom/kendex: Rust workspace; repo convention runs cargo test and cargo clippy. Code that writes kendex.toml or kendex.settings.toml must never reformat content inside TOML string values — flag any line-oriented pass over TOML text that is not string-state-aware; this has corrupted user…

vanillagreencom/kendex · 142 tokens