ariadne
Skill3 starsChanged 32 days agoMITView on GitHub
Use when initializing, diagnosing, running, recovering, or inspecting an Agent Toolkit Ariadne autonomous coding loop.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- mclaut/ariadnetheir own version, first seen Nov 1, 2026
- Kevin-Liu-01/Ariadnetheir own version, first seen Jun 5, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- Kevin-Liu-01/Ariadne★ 4same name
- mclaut/ariadne★ 3same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/general/code-reviewer/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- fastify-best-practicesSkill
- agent-toolkit-maintainerSkill
- onboard-repoSkill
- docker-patternsSkill
- accessibilitySkill
- code-reviewerSkill
- thermo-nuclear-code-quality-reviewSkill
- brag-slimSkill
- biome-formattingSkill
- instagram-public-researchSkill
- research-data-collectionSkill
- api-security-auditSkill
- cicd-security-auditSkill
- cloud-misconfiguration-auditSkill
- js-secrets-auditSkill
- llm-agent-security-auditSkill
- security-auditSkill
- source-leak-auditSkill
- ariadne-prdSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
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 39
Start in the configured Git branch with the user's existing work preserved.
- Very specific: line 40
Run doctor; fix reported state, runtime, or quality-check errors.
- Distinctive: line 42
Run normally only when the user accepts autonomous edits and commits.
- Distinctive: line 52
Before staging, Ariadne verifies that its lock, run coordinator, prompts, results, and raw process streams are still effectively ignored …
- Distinctive: line 61
If a runtime breaks any boundary, Ariadne fails safely and leaves the project diff available for deliberate recovery instead of risking an …
- Very specific: line 85
If a runtime changes the certified Git branch ref/HEAD, edits Ariadne's canonical PRD/progress files, or breaks a pinned operational …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
Ariadne
Ariadne is Agent Toolkit's runtime-neutral autonomous coding loop. The orchestrator owns state transitions, quality checks, and commits; the selected runtime agent implements one story at a time.
Commands
| Command | Purpose |
agent-toolkit ariadne init | Create or migrate .ariadne/ state, auto-detect available checks, and configure a runtime when unambiguous. |
agent-toolkit ariadne doctor | Diagnose Git, schema, lock, runtime, and quality-check readiness without running a story. |
agent-toolkit ariadne run | Autonomously select and execute stories, verify them, and let Ariadne commit successful work. |
agent-toolkit ariadne status | Report backlog counts, active story, branch, lock, latest run, and runtime health. |
Normal run is autonomous and mutates the repository. Use run --dry-run for
non-mutating inspection. Use --runtime <claude|codex|opencode|gemini|antigravity>
to override automatic runtime selection explicitly, --max-iterations <n> or
--max-runtime <duration> to bound a run, and --json for machine-readable
output. run first honors its explicit runtime, then a healthy configured
project runtime. An unverified, unavailable, or incompatible configured runtime
falls through to automatic selection: a single healthy candidate (or a sole
unverified candidate), a healthy global preference when set, an interactive
prompt when candidates remain ambiguous, or a non-interactive ambiguity error.
During init, Ariadne detects package quality checks automatically: it uses a
check script when present, otherwise it collects lint, typecheck, and
test scripts in that order. Repeated --check values replace those detected
checks. In a non-interactive project with no detected or explicit check, init
reports an error and requires --check.
Safe operation
- Start in the configured Git branch with the user's existing work preserved.
- Run
doctor; fix reported state, runtime, or quality-check errors. - Inspect
status, then userun --dry-runbefore a first autonomous run. - Run normally only when the user accepts autonomous edits and commits.
- Inspect
statusand.ariadne/runs/<run-id>/evidence after completion.
A runtime agent may edit project files and applicable AGENTS.md only for
genuinely reusable repository guidance. It must not create commits, push, or
edit .ariadne/prd.json or .ariadne/progress.md. Ariadne performs the
successful-story commit after its own acceptance and quality checks. Ariadne
never pushes and never resets,
reverts, cleans, checks out over changes, or discards a failed diff.
Before staging, Ariadne verifies that its lock, run coordinator, prompts,
results, and raw process streams are still effectively ignored and absent from
the tracked or staged index. It stages literal project paths, proves the full
project delta is represented, and publishes the exact certified tree on the
pre-runtime certified HEAD without running repository hooks. The state root,
runs root, run directory, lock coordinator, public lock, and operational files
must retain their captured filesystem identities. Output leaves must remain
present regular single-link files rather than symlinks or hard links. Ariadne
rechecks these boundaries immediately before staging and reference publication.
If a runtime breaks any boundary, Ariadne fails safely and leaves the project
diff available for deliberate recovery instead of risking an
operational-artifact commit.
Retry and recovery
Failed processes, results, criteria, and checks are recorded in .ariadne/runs/
and .ariadne/progress.md. Validated result and check evidence is summarized
there with secrets and raw process output removed. Ariadne preserves the story
diff and retries the same in_progress story. The default limit is three
attempts; exhaustion marks the story blocked and stops subsequent runs. A
blocked story continues to own its preserved dirty diff, so doctor reports
that state without misclassifying it as unrelated work. Diagnose with doctor
and the latest run artifacts, repair the underlying problem, then have the
operator deliberately reset the story to pending or in_progress before
resuming.
A commit failure is different: Ariadne records it, restores the story to
in_progress, and preserves the worktree diff, original shared-index state,
and run progress for manual resolution and a deliberate resume. It does not automatically retry or block a
commit failure, and it never hides failures or discards the preserved diff to
force progress. If Git publication already succeeded, a later machine-local
summary-write failure is structural and does not roll the completed story back.
If a runtime changes the certified Git branch ref/HEAD, edits Ariadne's
canonical PRD/progress files, or breaks a pinned operational boundary, Ariadne
persists the repository-root .ariadne-quarantine.json. The ignored
.ariadne-quarantine.checkpoint.json persists the last certified ref/HEAD
and exact PRD/progress certificates, while a canonical progress marker makes
that checkpoint mandatory after execution begins. Later init and run
invocations validate the earlier ownership boundary before probing or state
mutation and fail closed when the checkpoint is missing, malformed, or
mismatched; status and doctor remain available. Restore the recorded
certified ref/HEAD, canonical files, and operational layout, inspect the
preserved evidence, and only then deliberately remove the quarantine marker
before running again.
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.

