agentleFS
JeffKarlingJeffKarling/Anvil/.agents/skills/architect/SKILL.md

architect

Skill0 starsChanged 2 days agoMIT-0View on GitHub

Supervises the deterministic architecture lifecycle. Dispatches strategic_worker to partition physical targets and author contracts, validates with strategic_verifier, and coordinates tactical_worker and tactical_verifier waves across the DAG.

181 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ecosystem-reference, fractal-delegation and 10 more.

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.
  • naman10parikhSkill

    ecosystem-reference

    Quick-access directory of Claude Code ecosystem repos, skills, plugins, MCPs, and patterns. Use when improving the harness.

    Also in 5 projects that use this

  • naman10parikhSkill

    fractal-delegation

    Also in 5 projects that use this

  • tristanphvnSkill

    ketchup-video

    Catch the user up like /keel:ketchup, with a short silent video of slides for what happened and the waiting calls as text under it. Use for /keel:ketchup-video.

    Also in 14 projects that use this

  • jaydubya818Skill

    figure-it-out

    Use when novel or multi-phase work needs an auditable workflow and no narrower playbook fits.

    Also in 55 projects that use this

  • ianclearySkill

    principle-explain-the-number

    Apply before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.

    Also in 29 projects that use this

  • eryzerzSkill

    make-bot-ui

    Use when building a custom UI (page, dashboard, buttons) that should wake a Grok Bot over a webhook, when the user must provide a webhook sender key, or when exposing that UI on Tailscale.

    Also in 20 projects that use this

  • +6 more from naman10parikh

The file

We split this skill into the rules it gives an agent

12 rules. 0 are common, 6 are distinctive, and 6 are very specific to this project.

  • 0

    Common

    Most of the 169 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Run the manifest introspection with a clean working tree requirement before starting.

    line 13
  2. Distinctive:

    If the script exits non zero, refuse to proceed until all uncommitted files are committed or stashed.

    line 19
  3. Very specific:

    If an issue tracking ticket URL is provided, retrieve ticket descriptions or reproduction steps before delegating triage to …

    line 21
  4. Very specific:

    Anchor preservation: Never merge foundational anchors such as journal, perf_scope, or journal_signal_adapter.

    line 46
  5. Distinctive:

    Execution and verification parity: Never merge pure compute capabilities with stateful OS or I/O boundary capabilities.

    line 47
  6. Distinctive:

    Context locality and tactical helpers: Keep total worker context within $B_{\max} = 16384$ tokens.

    line 49
  7. Very specific:

    If the strategic verifier reports contract violations or gate failures, re-dispatch strategic_worker with the audit findings.

    line 71
  8. Very specific:

    Keep the monitor context free of private .cpp implementations and mechanical compiler loops.

    line 76
  9. Distinctive:

    Ensure tactical_worker and tactical_verifier are registered.

    line 78
  10. Very specific:

    When a tactical worker finishes, dispatch tactical_verifier via invoke_subagent with Workspace set to "inherit" and Model set to "flash" to …

    line 96
  11. Very specific:

    If tactical_verifier reports a clean pass, commit the component changes.

    line 97
  12. Distinctive:

    If an issue tracking ticket URL was provided with update instructions, post a completion summary with root cause findings, commit …

    line 111

“Common” is measured: kept by at least 6 in 10 of the 169 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Agentic DAG architecture workflow

You must act as the monitor for this workflow regardless of which model is active. The monitor is the deterministic supervisor and state machine governing the development lifecycle. The monitor never authors private C++ implementations or edits public contracts directly. Instead, the monitor coordinates the pipeline, delegates creation to strategic and tactical workers, and gates progress using independent verifiers.

Step 1. Verify clean Git state and dispatch strategic worker for triage

Run the manifest introspection with a clean working tree requirement before starting.

python3 .agents/skills/architect/scripts/sync_manifest.py --require-clean-git
  • If the script exits non zero, refuse to proceed until all uncommitted files are committed or stashed.
  • The monitor never analyzes defect root causes or inspects private implementations directly. When user defect feedback arrives, ticket URLs are provided, or requirements change, evaluate complexity and select the strategic_worker model tier (pro for complex defects or DAG mutations, flash for scoped anomalies).
  • If an issue tracking ticket URL is provided, retrieve ticket descriptions or reproduction steps before delegating triage to strategic_worker.
  • Dispatch strategic_worker with the defect report to investigate root causes, trace component data flows, determine whether public contracts or the DAG require modification, and produce an explicit tactical remediation plan with testability guidance.
  • Review cmake/dependencies.cmake and manifest.json before delegating work.
  • Inspect advisory telemetry notes under refactor_candidates in manifest.json. Pass relevant topology and contract insulation hints to the strategic worker so they are weighed against new requirements using deterministic net_gain and free rider neighborhood discounts.

Step 2. Dispatch strategic worker to partition capabilities and scaffold targets

Dispatch the strategic worker subagent to group capabilities into physical targets and author public contracts.

{
  "Subagents": [
    {
      "TypeName": "strategic_worker",
      "Role": "Strategic Architect",
      "Model": "pro",
      "Workspace": "inherit",
      "Prompt": "Translate requirements into physical targets under targets/ while weighing advisory refactor_candidates telemetry in production_artifacts/manifest.json. Fold positive net_gain refactors and free rider neighborhood optimizations into the topology. Scaffold target directories, author public headers in targets/<name>/<name>/<name>.hpp, author error catalogs in targets/<name>/<name>.catalog, configure targets/<name>/CMakeLists.txt, verify stubbed builds under Clang_Custom_Debug, and run sync_manifest.py --check. Report completion back to monitor."
    }
  ]
}

The strategic worker enforces five hard invariant gates:

  • Quotient DAG acyclicity. The physical target dependency graph must remain strictly acyclic.
  • Anchor preservation. Never merge foundational anchors such as journal, perf_scope, or journal_signal_adapter. Keep the app executable target isolated from library anchors.
  • Execution and verification parity. Never merge pure compute capabilities with stateful OS or I/O boundary capabilities. Pure compute targets must include benchmark suites.
  • Compiler regime homogeneity. Capabilities requiring custom flags must remain isolated. Keep targets requiring -fno-fast-math or explicit SIMD flags separated from standard targets.
  • Context locality and tactical helpers. Keep total worker context within $B_{\max} = 16384$ tokens. Single consumer helpers sharing the same compiler regime belong in private helper files inside the parent target.

Step 3. Audit contract anchor with strategic verifier and commit

Once the strategic worker finishes, verify the public contracts before committing the anchor.

  1. Dispatch the strategic verifier subagent via invoke_subagent.
{
  "Subagents": [
    {
      "TypeName": "strategic_verifier",
      "Role": "Strategic Verifier",
      "Model": "flash",
      "Workspace": "inherit",
      "Prompt": "Audit candidate public contracts and manifest gates. Verify that targets/<name>/<name>/<name>.hpp headers have zero quoted includes and zero third-party header leakage. Verify that python3 .agents/skills/architect/scripts/sync_manifest.py --check passes gates H1 through H5. Report a concise pass or fail verdict."
    }
  ]
}
  1. If the strategic verifier reports contract violations or gate failures, re-dispatch strategic_worker with the audit findings.
  2. Once the strategic verifier reports a clean pass, the monitor commits the contract anchor commit to the repository.

Step 4. Orchestrate tactical workers and verifiers in parallel

Keep the monitor context free of private .cpp implementations and mechanical compiler loops.

  1. Ensure tactical_worker and tactical_verifier are registered.
  2. Dispatch parallel waves of tactical workers using invoke_subagent with Workspace set to "share". Set Model from subagent_model in manifest.json using flash for $\Psi(C_i) < 0.70$ and pro for $\Psi(C_i) \ge 0.70$.
{
  "Subagents": [
    {
      "TypeName": "tactical_worker",
      "Role": "Component Implementer",
      "Model": "flash",
      "Workspace": "share",
      "Prompt": "Implement targets/<name>/<name>.cpp against targets/<name>/<name>/<name>.hpp and direct upstream headers. Verify unit tests, sanitizers, and baseline benchmark."
    }
  ]
}
  1. Restrict each tactical worker to reading and editing its assigned targets/<name>/ directory plus reading public upstream headers.
  2. When a tactical worker finishes, dispatch tactical_verifier via invoke_subagent with Workspace set to "inherit" and Model set to "flash" to audit the candidate diff.
  3. If tactical_verifier reports a clean pass, commit the component changes. If it detects test dilution or invariant evasion, discard the changes and dispatch a fresh worker with the audit findings.

Step 5. Whole graph verification and final release gate

Once all components merge cleanly, verify the integrated system across compilers and synchronize manifest.json.

./production_artifacts/build_all.sh custom debug
./production_artifacts/build_all.sh default release
python3 .agents/skills/architect/scripts/sync_manifest.py --check

Dispatch strategic_verifier via invoke_subagent to confirm all five invariant gates $H_1$ through $H_5$ pass and zero dirty benchmarks remain.

If an issue tracking ticket URL was provided with update instructions, post a completion summary with root cause findings, commit references, and benchmark diffs back to the tracking system.

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.