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/rcconsult/ppxai/claude-mdgit clone --depth 1 https://github.com/rcconsult/ppxaiWrote 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/instructions/rcconsult/ppxai/claude-md)<a href="https://agentmods.dev/instructions/rcconsult/ppxai/claude-md"><img src="https://agentmods.dev/badge/instructions/rcconsult/ppxai/claude-md.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 | $0.05461 | $0.05461 |
| Opus 5 | $0.02730 | $0.02730 |
| Sonnet 5 | $0.01092 | $0.01092 |
| Haiku 4.5 | $0.00546 | $0.00546 |
Grade A, and why
ppxai 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 5d 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 — 264 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.
Project Overview
ppxai is a terminal-based UI application for interacting with multiple AI providers (Perplexity, OpenAI, OpenRouter, local models). Interactive chat with model selection, conversation history, streaming, and AI-powered tools.
Current version: see pyproject.toml (single source of truth) or the latest release.
Release state: v1.18.8 and v1.18.7 are both released (2026-06-14 / 2026-06-13). v1.18.7 is the surface ppxai-sre's outlook-monitor agent consumes — the POST /v1/oneshot gateway shape (bearer auth) stays byte-identical. See docs/release-notes-v1.18.7.md and docs/archive/release-notes/release-notes-v1.18.6.md.
Active branch: feature/v1.19.0 (not yet released) — building the agent platform (ADR 0003 Stage 2): a durable, addressable /v1/agent/* run registry with a tool-capable sandboxed tier. Increments 1–9 are committed + live-trial-verified (run lifecycle, background exec, events/SSE monitor channel, AC-1 tool allowlist, AC-2 egress allowlist, budgets/cancel, spawn_subagent N=1, /v1/tokens + pluggable secret sources, per-run authz, AppState background_agents mirror), plus post-Inc-9 hardening §A–§K (provider-agnostic v1 tier, agent system-prompt framing, event-loop offload, egress SSRF guard, loopback UI auth exemption + /v1/agent/run carve-out, oneshot native web search, web /agentrun fire-and-forget). The interactive /task command family is now being built per docs/plan-task-command-sequencing.md (T1–T9, each live-trialable): T1 (web launch/observe — /task run·ls·show·watch·cancel), T2 (filesystem seal — tools.agent.sandbox read-path jail in ScopedToolManager, default-off), T3 (spec files — --spec <name> under sandbox.specs_dir, engine/agent_spec.py), and T4 (skills — --skill <name> under sandbox.skills_dir, engine/agent_skill.py; SKILL.md grant + references/ mounted into read-scope) are committed; T5 (interactive consent — waiting park + POST /runs/{id}/respond, consent card, state.json first write = debt (r)) is committed (trial recipe in the plan doc §T5); T6 (two-phase termination — completed_pending_ack hold + POST /runs/{id}/ack → finalized, Collect button, lazy retention reaper) is committed (trial recipe in the plan doc §T6); T7 (interrupted resume — POST /runs/{id}/resume conditional on the resume_refusal decision matrix, restart-orphan sweep at registry construction, Resume button; retires debt (r) — the state.json Triplet file now has producers T5/T6/T7 and its consumer) is committed (trial recipe in the plan doc §T7). T8 is split: T8a (VSCode port — taskController.ts with verb/endpoint/status parity sentinels against the web client, typed /v1/agent/* httpClient slice, consent QuickPick per VSCode idiom) is committed (trial recipe in the plan doc §T8a); T8b (TUI port) is ⏸️ PARKED (2026-07-07) pending a transport decision — the TUIs are in-process with no server channel, so either the registry+runner embed in-process (retires debt (t), recommended) or the TUIs grow an HTTP client; resume checklist in plan §T8. The /task family ships in web + VSCode; T9 (container tier-d) stays deferred. Design: docs/agent-task-command-design.html (/task surface + spec/skill files + sandbox config) and docs/agent-task-lifecycle.html (run-state machine, /respond·/ack·/resume). See docs/plan-v1.19.0-sequencing.md for the Stage-2 increment plan + build contract, docs/agent-platform-call-graphs.md for per-increment route→event call graphs (§A–§K = post-Inc-9 fixes; T1–T2 client + seal, T5–T7 lifecycle, and T8a VSCode-port graphs appended), and docs/decisions/0003-agent-platform-architecture.md for the architecture. Open deferred work lives in the rolling docs/debt-inventory.md (per-version snapshots archived under docs/archive/).
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.
- 5d ago First seen · 264 lines · 5,461 tokens per session scan A 20639a7ec6b3
ppxai CLAUDE.md is an instructions file published in the GitHub repository rcconsult/ppxai (5 stars, last pushed 3d ago), licensed MIT. It adds 5,461 tokens to every session, about $0.0273 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
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).