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/capabilityhostprotocol/chp-core/agents-mdgit clone --depth 1 https://github.com/capabilityhostprotocol/chp-coreWhat 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 | $0.00775 | $0.00775 |
| Opus 5 | $0.00387 | $0.00387 |
| Sonnet 5 | $0.00155 | $0.00155 |
| Haiku 4.5 | $0.00077 | $0.00077 |
Grade A, and why
chp-core 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 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.
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 — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Capability Host Protocol
CHP is a protocol and Python SDK for making agent, tool, and system execution observable, replayable, and governable. Every function wrapped as a @capability gets automatic evidence emission, correlation propagation, replay by session ID, and optional policy enforcement — with zero mandatory infrastructure. The reference host is LocalCapabilityHost in packages/python/chp_core/.
Three invariants you must never violate
- Evidence is append-only. Never modify or delete rows in an evidence SQLite store. The SHA256 hash chain breaks if any row changes.
store.pyis insert-only by design. - Preserve caller correlation IDs. If an
InvocationEnvelopearrives with acorrelation_id, the host must forward it verbatim into every evidence event and the result. Never generate a new ID over a supplied one. - Spec, schemas, and types must stay in sync. Any change to
spec/chp-v0.1.md,schemas/*.json, orpackages/python/chp_core/types.pymust be validated withchp work check-alignment --repo-root .(runs 41 cross-artifact checks). CI will catch drift, but run it locally first.
Key commands
# Fast test suite (~6s)
cd packages/python && python -m pytest tests/ -m "not slow" -q --no-cov
# Full test suite
cd packages/python && python -m pytest tests/ -q --no-cov
# Protocol conformance (9 checks)
python conformance/runner.py
# Spec/schema/type alignment (41 checks)
PYTHONPATH=packages/python chp work check-alignment --repo-root .
# Wire evidence capture for every Claude Code session
PYTHONPATH=packages/python chp hooks install
Navigation
| Where to look | What you'll find |
|---|---|
spec/chp-v0.1.md |
Normative protocol — start here for definitions and MUST/SHOULD requirements |
schemas/ |
JSON Schema for every protocol object (29 files) |
packages/python/chp_core/host.py |
LocalCapabilityHost — registration, invocation, evidence emission |
packages/python/chp_core/store.py |
SQLiteEvidenceStore — append-only, SHA256-chained |
packages/python/chp_core/types.py |
Python dataclasses for all protocol objects |
conformance/runner.py |
9 conformance checks against a live host |
docs/adopter-quickstart.md |
10-minute path to first evidence event |
examples/ |
14 runnable demos |
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 · 49 lines · 775 tokens per session scan A 09b7bb264227
chp-core AGENTS.md is an instructions file published in the GitHub repository capabilityhostprotocol/chp-core (1 stars, last pushed 5d ago), licensed Apache-2.0. It adds 775 tokens to every session, about $0.0039 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
lmnr CLAUDE.md
Instructions for lmnr-ai/lmnr, covering claude.md, project overview, repository structure, development commands and frontend (next.js).
cruxible AGENTS.md
AGENTS.md instructions for cruxible-ai/cruxible, covering agents.md, project overview, commands, install dependencies and run tests.
rungraph CLAUDE.md
Instructions for fayzan123/rungraph, covering rungraph — project instructions, current state, the one loop, non-negotiable constraints (from the approved spec) and shared code, one implementation.
symbi-gemini-cli GEMINI.md
Instructions for ThirdKeyAI/symbi-gemini-cli, covering symbiont governance context, key concepts, available mcp tools (via symbi server), file conventions and dual-mode operation.
symbi-gemini-cli GEMINI.md
Gemini CLI instructions for Ezee234/symbi-gemini-cli, covering symbiont governance context, key concepts, available mcp tools (via symbi server), file conventions and dual-mode operation.
openstatus AGENTS.md
AGENTS.md instructions for openstatusHQ/openstatus, covering agents.md, verify your change, toolchain, architecture and tests.