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/pmbstyle/octopal/agents-mdgit clone --depth 1 https://github.com/pmbstyle/OctopalWrote 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/pmbstyle/octopal/agents-md)<a href="https://agentmods.dev/instructions/pmbstyle/octopal/agents-md"><img src="https://agentmods.dev/badge/instructions/pmbstyle/octopal/agents-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 5.1 | $0.02634 | $0.02634 |
| Opus 5 | $0.01317 | $0.01317 |
| Sonnet 5 | $0.00527 | $0.00527 |
| Haiku 4.5 | $0.00263 | $0.00263 |
Grade A, and why
Octopal 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 6d 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 — 93 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repository Guidelines
Project Structure & Module Organization
src/octopal/contains the main Python package: CLI, channels, gateway, memory, policy, providers, Octo runtime, scheduler, WhatsApp/Telegram integrations, workers, and shared utilities.webapp/holds the Vite-based dashboard frontend.src/contains app code anddist/contains built assets.tests/contains the pytest suite for CLI, dashboard, runtime, worker orchestration, memory, and channel behavior.scripts/contains setup and maintenance helpers such as bootstrap, repository-hosted install scripts (octopal.sh,octopal.ps1), and worker-template sync utilities.docker/contains container assets, including the worker image Dockerfile.data/is runtime state storage for SQLite, metrics, auth state, and logs; avoid committing generated contents.workspace/is the default Octo/worker workspace and scratch area.workspace_templates/contains bootstrap content copied into new workspaces.docs/stores additional project documentation.
Build, Test, and Development Commands
uv syncinstalls the project and the cross-platform CPU-only ONNX semantic-memory runtime without PyTorch; the pinned local model bundle downloads and validates on first runtime start.python -m venv .venvandpip install -e .[dev]are the non-uveditable setup path.uv run octopal configureruns the interactive configuration wizard and bootstraps missing workspace files.uv run octopal config show [--reveal-secrets]prints the effective config for inspection, anduv run octopal config migratewrites the current.env-backed settings intoconfig.json.uv run octopal startstarts Octopal in background mode.uv run octopal start --foregroundruns the Octo and gateway in the foreground.uv run octopal stop,uv run octopal restart, anduv run octopal statusmanage the local runtime.uv run octopal updateapplies the latest release update flow for an existing install.uv run octopal logsprintsdata/logs/octopal.log, anduv run octopal logs --followtails it.uv run octopal gatewaystarts the FastAPI gateway directly.uv run octopal dashboard --onceprints one dashboard snapshot;uv run octopal dashboard --watchruns the live terminal dashboard;uv run octopal dashboard --jsonemits a machine-readable snapshot;--compactand--last <N>tune terminal output.uv run octopal workers listshows registered workers, their current status, and the current task when available.uv run octopal audit list [--limit <N>]lists recent audit events, anduv run octopal audit show <event-id>prints one audit event in detail.uv run octopal episodes list [--worker-run-id <id>] [--limit <N>] [--json]anduv run octopal episodes show <episode-id> [--json]inspect metadata-only execution evidence.--reveal-evidenceexplicitly decrypts raw content,uv run octopal episodes erase-evidence <episode-id> [--yes]securely deletes only the encrypted raw payload, anduv run octopal episodes purge-storageretries a blocked WAL purge without deleting live evidence.uv run octopal connector status [--json]checks connector authorization/readiness after connector setup flows;uv run octopal connector auth <name>anduv run octopal connector disconnect <name> [--forget-credentials]handle CLI-based connector auth maintenance.uv run octopal sync-worker-templates --overwriterefreshes default worker templates intoworkspace/workers.uv run octopal memory statsanduv run octopal memory cleanup [--keep-days <days>] [--keep-count <count>] [--dry-run]cover common memory maintenance.uv run octopal memory proposals list [--state <state>] [--json]andshow <proposal-id> [--json]inspect provenance-bearing canon proposals;promote <proposal-id> [--yes]activates one, whiledeprecate <proposal-id> [--yes]rolls back its active canon content.uv run octopal memory recipes propose <definition.json> [--include-matching]records an episode-backed candidate;context <recipe-id> --jsonexports bounded benchmark context,evaluate <recipe-id> <baseline.json> <candidate.json>records held-out evidence, andlist,show,promote <recipe-id> --yes, anddeprecate <recipe-id> --yesmanage its explicit lifecycle.uv run octopal whatsapp install-bridge,uv run octopal whatsapp link [--timeout <seconds>],uv run octopal whatsapp status, anduv run octopal whatsapp logoutmanage the WhatsApp bridge lifecycle.uv run octopal tools resolve [--profile <name>] [--preset all|octo] [--blocked|--available-only] [--json]explains which tools are available and why others are blocked.uv run octopal skill list [--json],uv run octopal skill install <source>,uv run octopal skill update <skill-id>,uv run octopal skill verify <skill-id>,uv run octopal skill enable <skill-id>,uv run octopal skill disable <skill-id>,uv run octopal skill trust <skill-id> [--force],uv run octopal skill untrust <skill-id>,uv run octopal skill prepare-env <skill-id>,uv run octopal skill remove-env <skill-id>, anduv run octopal skill remove <skill-id>cover the installed skill workflow.uv run octopal build-worker-image --tag octopal-worker:latestbuilds the Docker worker image.uv run pytestruns the test suite.uv run ruff check .,uv run black --check ., anduv run mypyare the configured lint/format/type-check commands.npm installandnpm run buildfromwebapp/build the dashboard bundle manually when needed.uv run python scripts/benchmark_web_backends.py --backends basic,markdown_new,webclaw,browser --webclaw-binary <path> --repeat 3 --output data/web-backend-benchmark.jsoncompares web extraction backends without starting an Octopal runtime.uv run python scripts/worker_bench.py --suite <suite.json> --workspace workspace_templates --mode live --config <test-config.json> --preflight-onlyvalidates provider budgets, read-only tools, credentials, and per-run isolation without making provider or tool calls. Remove--preflight-onlyonly for a manually approved bounded live run.uv run python scripts/worker_bench.py --suite <suite.json> --mode replay --replay-dir <artifacts-dir> [--baseline <result.json>] --out <result.json>grades saved worker JSONL offline; baseline mode exits non-zero only for outcome regressions. Live suites must explicitly declare provider/model, strict thinking, LLM-call and tool-call caps, token/cost ceilings, and current model pricing.uv run python scripts/tool_call_bench.py --suite <suite.json> --config <test-config.json> --preflight-only [--out <result.json>]validates a zero-execution tool-argument benchmark. Remove--preflight-onlyonly for a bounded live run; suites are capped at three single-attempt model calls and artifacts retain no prompts or arguments.uv run octopal adaptation cluster <bench-result.json>,uv run octopal adaptation propose <definition.json>,uv run octopal adaptation context <candidate-id>, anduv run octopal adaptation evaluate <candidate-id> <baseline.json> <candidate.json>cover the offline adaptation workflow.promote <candidate-id> --yesandrollback <candidate-id> --yesupdate only the approved registry version; runtime prompts, tools, routes, and recipes remain untouched.bandit-readinessreports why online routing is still disabled.
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.
- 6d ago First seen · 93 lines · 2,634 tokens per session scan A f1eb8bf7603e
Octopal AGENTS.md is an instructions file published in the GitHub repository pmbstyle/Octopal (10 stars, last pushed 20d ago), licensed MIT. It adds 2,634 tokens to every session, about $0.0132 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
iClaw AGENTS.md
AGENTS.md instructions for iClawApp/iClaw, covering what iclaw is, what iclaw is not, architecture you need to know, runtime modes (work / safe work / incognito) and the canonical client paths.
clawlet AGENTS.md
Instructions for mosaxiv/clawlet, covering repository guidelines, project structure, build, test, and development commands, build a local binary and run from source.
librefang AGENTS.md
AGENTS.md instructions for librefang/librefang, covering stack, layout, build, architecture and kernelhandle trait.
autonomous-os AGENTS.md
AGENTS.md instructions for autonomous-ai/autonomous-os, covering agents.md, multi-ide rules, working style, parallel work / subagents and device access rules.
openclaw CLAUDE.md
Claude Code instructions for openclaw/openclaw, a project described as: The AI that really does things. Any OS. Any Platform. The lobster way. 🦞.
souls-directory CLAUDE.md
Instructions for thedaviddias/souls-directory, covering claude.md — project context for ai assistants, project overview, tech stack, critical rules and database is convex.