agentleFS
Sign inStart free
onelastcommitonelastcommit/noctra/.claude/skills/naming/SKILL.md

naming

Skill4 starsChanged 29 days agoMITView on GitHub

Use when changing how Noctra names or labels what it creates on GitHub or disk: branch names, ticket and sweep identifiers, worktree or log paths, commit messages, PR titles and footers, the model label, or the hidden markers the PR watcher matches.

16 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use testing.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • NVIDIASkill

    testing

    Run Python and Rust tests for physicsnemo-curator using uv + pytest with coverage reporting via pytest-cov, and cargo-nextest for Rust tests. Includes benchmark workflows with pytest-benchmark and criterion.

    Also in 3 projects that use this

The file

Create a free account to mount
Files
SKILL.md

Naming in Noctra

Every name Noctra writes is read back later by something else: a branch name by the PR watcher, a footer by the lessons extractor, an identifier by the active set. Treat each one as a round trip. A change to how a name is written is only finished when every reader listed below recognises the new form.

Branches

KindFormWritten byRead back by
Ticketnoctra/<identifier, lowercased>, e.g. noctra/eng-460repo.BranchNamegithub.ListNoctraPRs (prefix), identifierFromBranch (pipeline/iterate.go), cleanup (prefix)
Sweepnoctra/sweep-<task>, e.g. noctra/sweep-deps-updatesweep.SweepBranchNamesweep.TaskSuffixFromBranch, identifierFromBranch, the open-PR check in processSweepTask
  • The repo is left out of sweep branch names because a branch already lives in its repo. It stays in the identifier (below).
  • noctra/ is reserved for branches Noctra creates. The rule for humans and other agents is in CLAUDE.md.
  • github.ListNoctraPRs claims a PR only when all three match: the noctra/ prefix, a body marker (below), and the author: app/<slug> in GitHub App mode, @me in personal-token mode (github.Client.Author).

Identifiers, directories and logs

  • Ticket identifier: the tracker's ID, e.g. ENG-460 or GH-<SLUG>-<n>.
  • Sweep identifier: SWEEP-<repo-slug>-<task> (sweep.SweepIdentifier). It must stay repo-qualified: it keys the active set, the worktree directory and the log file, so two repos running the same task would otherwise block each other. identifierFromBranch rebuilds it from a short sweep branch plus the PR's owner/name.
  • Repo slug: repo.Slug(owner/name), e.g. onelastcommit-onenote-mcp. It equals the clone directory under ~/.noctra-repos, which is where the scheduler reads it (repo.SlugFromPath), so both sides agree.
  • Worktree ~/.noctra-worktrees/<identifier>; log logs/<identifier>.log under the config dir.
  • Show people owner/name, never the slug: the sweep PR body's Repo: line uses sweepRepoName.

Commit messages and PR titles

  • Subject and PR title follow Conventional Commits when the target repo uses them (repo.UsesConventionalCommits). The type comes from the agent's release-bump suggestion (patch gives fix, minor gives feat, major gives feat! plus BREAKING CHANGE), built by conventionalSubject. Otherwise the subject is <id>: <title>.
  • The body opens with one of three fixed phrases: Implemented by Noctra, Follow-up commit by Noctra, Autonomous maintenance by Noctra. lessons.noctraCommitRe matches them to keep Noctra from learning from its own commits, so the phrases stay word for word; anything may follow them.
  • A Co-authored-by: trailer names the agent backend (Backend.CoAuthor).

PR bodies and replies

  • Footer: *<phrase> by [Noctra](https://github.com/onelastcommit/noctra) 🦉 using <runner label>*.
  • github.NoctraPRBodyMarker (<!-- noctra-authored -->) goes in every PR body Noctra writes, ticket, sweep or salvaged draft alike. The watcher also accepts the legacy visible footer by [Noctra] for PRs that predate the marker.
  • github.NoctraReplyMarker (<!-- noctra-reply -->) goes in every comment and thread reply Noctra posts, including Gemini's inline findings. Without it the watcher re-reads Noctra's own words as review feedback and loops.

The model label

agent.RunnerLabel(backend.Label(), usage.Model) produces the runner label, e.g. Claude Code (Opus 5.5) or OpenAI Codex (GPT-5.5). With no model it is the bare backend label. agent.ModelDisplayName prettifies claude-<family>-<major>[-<minor>][-<date>] IDs (dropping a [1m]-style suffix) and gpt-* IDs, and shows anything else verbatim.

BackendWhere the model comes from
Claude CodemodelUsage in the run's JSON result, taking the costliest model since smaller ones handle sub-tasks. In the streaming path (token-capped runs) an aborted run never gets a result, so it falls back to the assistant messages seen, weighted by output tokens.
CodexThe session header line model: <id> that codex exec prints before the transcript (codexModel).
Copilot, AntigravityNot read yet. Both accept --model, but neither output format has been checked against a real run, so their label stays bare.

Parse a backend's model only from output you have captured from a real run. A guessed format fails silently and shows nothing or the wrong model.

Bot identity

In GitHub App mode commits are authored and committed as noctra-agent[bot] <336615789+noctra-agent[bot]@users.noreply.github.com>, taken from the linked instance (ghauth.IdentityEnv). The slug, bot login and bot ID come from the token service's /config, so a self-hosted app gets its own names with no code change.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free