selftune AGENTS.md

Repository instructions for Selftune, a TypeScript and Bun project that observes how AI agents are used and updates their skill descriptions to fit real work.

In plain words
What is it for?
Working on Selftune's agent-first architecture, CLI, local dashboard, databases, observability data, skill-learning behaviour, and Effect-based runtime code.
Why use it?
They explain the repository structure, its separate data systems, and rules for using the bundled Effect reference without editing it or importing from it.

Instructions file for CodexOpenCode

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/selftune-dev/selftune/agents-md
Clone the repo
git clone --depth 1 https://github.com/selftune-dev/selftune

Made for: Codex, OpenCode.

Per session 3,409 This file is loaded in full into every session.
When invoked 3,409 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 $0.03409 $0.03409
Opus 5 $0.01705 $0.01705
Sonnet 5 $0.00682 $0.00682
Haiku 4.5 $0.00341 $0.00341

Measured 2d ago against content hash 7929c6ceef88, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

selftune 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 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.

AGENTS.md · 196 lines

How it starts

The opening of the file, as written. The whole thing — 196 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

Repository Overview

selftune — Self-improving skills for AI agents. Watches real sessions, learns how users actually work, and evolves skill descriptions to match. Supports Claude Code, Codex, OpenCode, OpenClaw, and Pi.

Stack: TypeScript on Bun for the CLI, Drizzle over Bun SQLite for operational/product state, a separate DuckDB observability-analytics domain, legacy/export JSONL recovery paths, Effect for owned runtime lifecycles, and a local React/Vite dashboard SPA.

Vendored Effect Reference

  • .repos/effect is the pinned Effect 4 source used as read-only reference material.
  • Do not edit or import from .repos/effect; application code must continue importing normal package dependencies.
  • Before writing Effect code, read .repos/effect/LLMS.md and inspect matching source, tests, and API signatures.
  • Editor search, file watching, and auto-imports must remain excluded from .repos/**.

Agent-First Architecture

selftune is a skill consumed by AI agents, not a CLI tool used by humans directly.

The user's interaction model is:

  1. Install the skill: npx skills add selftune-dev/selftune
  2. Tell their agent: "set up selftune" / "improve my skills" / "how are my skills doing?"
  3. The agent reads skill/SKILL.md, routes to the correct workflow, and runs CLI commands

The selftune binary composed in apps/cli/ is the agent's API. The skill definition (skill/SKILL.md) is the product surface. Workflow docs (skill/workflows/) are the agent's instruction manual. Users rarely if ever run selftune commands directly — their coding agent does it for them.

When developing selftune:

  • Changes to CLI behavior must be reflected in the corresponding skill/workflows/*.md doc
  • New CLI commands need a workflow doc and a routing entry in skill/SKILL.md
  • Error messages should guide the agent, not the human (e.g., suggest the next CLI command, not "check the docs")
  • The SKILL.md routing table and trigger keywords are as important as the CLI code itself — they determine whether the agent can find and use the feature
  • skill/SKILL.md and skill/workflows/*.md are shipped product surface for users' agents. Do not put repo-local contributor commands or monorepo workflow there (for example: cd oss/selftune, bun run dev, Vite/HMR details). Put contributor guidance in README.md, AGENTS.md, or internal docs instead.

Read the full file on GitHub · 196 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 · 196 lines · 3,409 tokens per session scan A 7929c6ceef88

Subscribe to this mod's changes

selftune AGENTS.md is an instructions file published in the GitHub repository selftune-dev/selftune (16 stars, last pushed 2d ago), licensed MIT. It adds 3,409 tokens to every session, about $0.0170 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.