pohunek
AGENTS.md8 starsChanged 9 days agoMITView on GitHub
Canonical repository entry point. Keep this file short: route each task to the current authoritative instructions instead of repeating them. When you change a command, a crate boundary, or a convention, update the guide that owns it (this file, docs/development.md, or a scoped guide) in the same change. Repository text, code, comments, and commit messages are in English. Security takes priority over quality, simplicity, and speed. Follow the user's authorized scope; never infer permission to publish, merge, release, deploy, or…
- Reads credentials
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,439 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository13 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- ms-rustSkill
- contributeSkill
- coordinateSkill
- deliver-issueSkill
- gatesSkill
- github-workflowSkill
- merge-advanceSkill
- milestone-reviewSkill
- milestoneSkill
- plan-phaseSkill
- pr-handoffSkill
- releaseSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
Read the linked guide when its trigger applies.
- Very specific: line 44
A new plugin (runtime package or workflow plugin): never add it to this repository.
- Very specific: line 55
Rust edits: before touching .rs, read the relevant files named by .agents/rust-guidelines/SKILL.md, starting with …
- Distinctive: line 58
Gate execution: use the gates skill, which follows the verification guide above.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md
Canonical repository entry point. Keep this file short: route each task to the
current authoritative instructions instead of repeating them. When you change a
command, a crate boundary, or a convention, update the guide that owns it (this
file, docs/development.md, or a scoped guide) in the same change. Repository text,
code, comments, and commit messages are in English. Security takes priority over
quality, simplicity, and speed. Follow the user's authorized scope; never infer
permission to publish, merge, release, deploy, or use credentials.
For every task, including read-only inspection, diagnosis, and review, never
read .env*, key or certificate files, or disclose token values. The detailed
secret-handling policy lives in
.agents/instructions/coding.md.
About this repository
Pohunek is an owner-first control plane for durable coding-agent sessions. The Rust daemon owns session state, isolated workers own live PTYs, and the CLI and versioned SDKs are the public client boundary. The core ships no UI; team relay is additive and not implied to be shipped. See the authoritative architecture and public API.
Required instruction routes
Read the linked guide when its trigger applies; a task does not require reading unrelated runbooks.
- Every product or architecture change: .agents/instructions/project.md
for the compatibility window, persisted-state rules, and hard product
boundaries.
docs/architecture.mdis authoritative when design details differ. - Any code, test, documentation, build, or validation change: .agents/instructions/verification.md for CI gates, fast loops, testing policy, hermeticity, and evidence requirements.
- Any implementation, wire-protocol, CLI, or public contract change
(including edits to
docs/public-api.md,docs/cli.md, ordocs/knowledge/): .agents/instructions/coding.md for Rust, security, configuration, protocol ripple, knowledge, CLI completion, and relay constraints. - Issue, project, branch, commit, pull request, merge, or release work:
.agents/instructions/workflow.md and the
github-workflow skill. That skill
owns issue/project operations;
.github/agent-workflow.jsonis its config. - A new plugin (runtime package or workflow plugin): never add it to this
repository. An official plugin lives in its own repository
pohunekd/plugin-<name>, created frompohunekd/pohunek-plugin-templateby the owner; a third-party plugin lives in its author's account. A plugin becomes installable by id through a pull request topohunekd/pohunek-plugin-index, never through a change here. Change this repository only for a core capability the plugin needs. Naming, required CI, compatibility and admission are in Plugin repositories. The template, kit, index and snapshot are available from a later release (see "What is available when" there); until then no new plugin is added here. - Rust edits: before touching
.rs, read the relevant files named by .agents/rust-guidelines/SKILL.md, starting with11_universal_guidelines.md. - Gate execution: use the gates skill, which follows the verification guide above.
- macOS native tests: the
platform contracts (arm64)lane runs from one nextest archive, andmacOS tier packages (arm64)runs on main and nightly only; both are described in the verification guide's CI tiers section.
For workspace layout and routine development commands, see
docs/development.md. The canonical skills live under
.agents/skills/; .claude/skills/ contains discovery links only.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

