agentleFS
Sign inStart free
DshamirDshamir/slm-forge/skills/forge-architect/SKILL.md

forge-architect

Skill1 starsChanged 6 months agoMITView on GitHub

Selects base model + training regime + tokenizer strategy + chat template from config/whitelist.json by filtering on license preference and parameter ceiling. Writes manifest.plan with rationale. Bridges from v2 plan.json when invoked under the v2-bridged dispatch path.

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    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
  1. Very specific:

    Write manifest.plan.* with a rationale string + selected_at

    line 21
  2. Very specific:

    If manifest.plan is already populated, exits 0 (no-op).

    line 35

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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.

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

forge-architect

When this fires

Phase position: ARCHITECT — after forge-intake, before forge-estimate. On v2-bridged runs, reads plan.json (already produced by forge-plan) and copies the architectural fields into the legacy manifest.

What it does

  1. Load config/whitelist.json (the whitelisted base models)
  2. Filter by manifest.spec.license_preference and the v1 size ceiling
  3. Score remaining candidates by domain fit + tokenizer coverage + size
  4. Pick the best base model + regime (lora-sft / qlora-sft / full-sft)
  5. Pick the chat template + tokenizer strategy that go with that base
  6. Write manifest.plan.* with a rationale string + selected_at
  7. Advance state to ESTIMATE

Inputs

  • $1 = forge-id (legacy) or v2-<run-id> (v2 bridge)
  • manifest.spec (license_preference, language, constraints)
  • config/whitelist.json (base model whitelist)
  • plan.json if the run was bridged from v2 (overrides whitelist scoring)

Outputs

  • manifest.plan.{base_model, regime, target_params, framework, chat_template, tokenizer_strategy, rationale, selected_at}
  • manifest.state.current_phase = ESTIMATE

Idempotency

If manifest.plan is already populated, exits 0 (no-op).

Failure modes

ExitReason
0plan written or already-set
1spec.license_preference incompatible with every whitelisted model
64no forge-id provided

External resources

  • Local disk only — reads config/whitelist.json, writes manifest

Cost class

free — deterministic config selection, no API calls.

Depends on

  • forge-intake (populated manifest.spec)
  • forge-plan (when v2-bridged — supplies plan.json)

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.

Reviews can't be read right now.

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