Revfactory Harness operator
When to use this skill
Use it for the revfactory/harness product: install/inspect the plugin, construct or
extend a domain team, audit generated files, migrate v1 artifacts, or reflect feedback
through its evolve skill. Choose one operation before changing a project.
This catalog's revfactory-harness is an operator guide, not the upstream harness
plugin. The old catalog name harness remains retired in favor of harness2; do not
restore that directory, overwrite a foreign harness, or alias the two projects.
Generic Propose–Probe–Compose work stays with harness2.
Instructions
- Run the offline gate. From this repository root:
python3 .agent-skills/revfactory-harness/scripts/preflight.py --json
From an installed skill, use python3 "$SKILLS_ROOT/revfactory-harness/scripts/preflight.py" --json.
Add --repo /path/to/harness --expect-commit 92d9f1b29e7af81acebf51a9fcd1832e739a1414
to verify the checkout and all pinned public-file hashes. Add --project /path/to/project
to inspect target-path types and direct-copy collisions. The helper never runs Claude,
plugin installation, upstream scripts, models, or network requests; it never reads credentials.
- Separate catalog installation from plugin activation. Read
installation and provenance. Installing this guide copies
documents and helpers only. Preview the router install; obtain approval for a clone,
marketplace registration, plugin install, or writes to
.claude/ and CLAUDE.md.
Prefer a verified pinned checkout as the local marketplace. Never silently copy over
.claude/skills/harness, harness-evolve, or evolve, including linked/shared roots.
- Audit before generating. Read factory and evolution workflow.
Inspect
.claude/agents/, .claude/skills/, CLAUDE.md, and existing orchestration.
Report new-build, extension, maintenance, or v1 migration scope. Reuse existing roles
and skills instead of creating same-purpose duplicates. Preserve user edits and output language.
- Choose a mode from the control flow and actual host capabilities. Deterministic
lists/loops use the upstream Workflow mode only if
Workflow is exposed; repeated
expert negotiation uses named agents plus messaging/tasks only if those tools are
present; one-shot independent work uses available subagents. Check the actual tools,
supported parameters, models, and budget before execution. Upstream documentation is
not proof that this session supports Workflow, Agent(name:), or fable. If a
required capability is absent, report blocked; do not invent APIs or silently fall back.
- Generate the smallest approved team. Choose one of pipeline, fan-out/fan-in,
expert pool, producer-reviewer, supervisor, or bounded hierarchical delegation.
Define reusable roles under
.claude/agents/<name>.md; methods and exactly one
orchestrator under .claude/skills/<name>/SKILL.md. Specify inputs, outputs, schemas,
ownership, failure/missing-result policy, and measurable completion. Keep CLAUDE.md
to trigger pointers and change history. Read-only reviewers get no Write/Edit tools.
- Keep execution and feedback bounded. Start with a small sample and explicit token,
iteration, and concurrency limits. Provider calls and generated commands retain host
approval policies. Treat failed, missing, or
uncertain outputs as unresolved, not
passes. Freeze producer artifacts at phase boundaries before consumers use them.
For feedback on an existing team, use plugin /harness:evolve (or direct-copy
/harness-evolve), change one cause at a time, and record before/after evidence.
Structural redesign returns to the factory; evolve does not silently add/remove roles.
- Verify the generated result, not merely file presence. Check frontmatter,
cross-references, actual runtime tools, producer/consumer contracts, and trigger cases.
Remove obsolete v1 constructs only after confirming the target runtime's replacement
capabilities. Run a scoped task; compare with/without-skill results only with execution
approval and report real artifacts, errors, timing/tokens when supplied, and unscored
failures. A passed offline preflight does not prove plugin loading or team quality.
Examples
Inspect a pinned checkout and project without execution
python3 .agent-skills/revfactory-harness/scripts/preflight.py --json \
--repo /path/to/harness \
--expect-commit 92d9f1b29e7af81acebf51a9fcd1832e739a1414 \
--project /path/to/project
Build or evolve after the approved plugin install
In the selected Claude Code session, request revfactory/harness로 문헌 검토 팀을 구성해줘.
After a real run, invoke /harness:evolve with the failed output and requested correction.
Keep the initial team, observed delta, adopted change, and the new verification result.
Best practices
- Progressive disclosure: keep this front door small; load only the relevant pinned
upstream reference linked from the local workflow guide.
- Select available models by task difficulty and cost; never assign every role the
most expensive tier or assume an upstream alias resolves on another host.
- Do not enable
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS as an automatic migration fix.
- Preserve
_workspace/ evidence and run IDs; do not recreate missing receipts.
- Upstream's v1 self-reported A/B gains are not evidence for this v2 integration.
References
- Pinned provenance, installation modes, and host boundaries
- Factory, migration, quality, and evolution workflow
- Audited public-file SHA-256 manifest
- Behavioral routing and approval cases
1---2name: revfactory-harness3description: >4 Operate revfactory/harness, the Apache-2.0 Claude Code team-architecture factory:5 audit or install its plugin, design domain-specific agents and skills, select an6 execution mode, migrate v1 artifacts, and evolve an existing generated harness.7 Use when the request names revfactory/harness, Harness v2, harness-marketplace,8 /harness:evolve, or its Korean team-factory workflow. Triggers on: 하네스 구성,9 하네스 회고, revfactory harness, harness factory, harness migration.10 Do not use for google-research/harness2 benchmarks or generic harness optimization11 (`harness2`), standalone skill authoring (`skill-standardization`), or running a12 Ruflo swarm (`ruflo`).13allowed-tools: Bash Read Write Edit Glob Grep14compatibility: >15 Python 3.10+ for the offline helper. Upstream execution requires an installed,16 authenticated Claude Code with the selected tools and model aliases actually17 available. Other hosts can inspect this guide but do not gain Claude-native tools.18license: Apache-2.019metadata:20 version: "1.0.0"21 source: https://skillmd.com/skills/akillness/revfactory-harness22 upstream: https://github.com/revfactory/harness23 upstream_commit: 92d9f1b29e7af81acebf51a9fcd1832e739a141424 upstream_version: "2.1.0"25---2627# Revfactory Harness operator2829## When to use this skill3031Use it for the **revfactory/harness** product: install/inspect the plugin, construct or32extend a domain team, audit generated files, migrate v1 artifacts, or reflect feedback33through its evolve skill. Choose one operation before changing a project.3435This catalog's `revfactory-harness` is an operator guide, not the upstream `harness`36plugin. The old catalog name `harness` remains retired in favor of `harness2`; do not37restore that directory, overwrite a foreign `harness`, or alias the two projects.38Generic Propose–Probe–Compose work stays with `harness2`.3940## Instructions41421. **Run the offline gate.** From this repository root:43 ```bash44 python3 .agent-skills/revfactory-harness/scripts/preflight.py --json45 ```46 From an installed skill, use `python3 "$SKILLS_ROOT/revfactory-harness/scripts/preflight.py" --json`.47 Add `--repo /path/to/harness --expect-commit 92d9f1b29e7af81acebf51a9fcd1832e739a1414`48 to verify the checkout and all pinned public-file hashes. Add `--project /path/to/project`49 to inspect target-path types and direct-copy collisions. The helper never runs Claude,50 plugin installation, upstream scripts, models, or network requests; it never reads credentials.512. **Separate catalog installation from plugin activation.** Read52 [installation and provenance](references/upstream.md). Installing this guide copies53 documents and helpers only. Preview the router install; obtain approval for a clone,54 marketplace registration, plugin install, or writes to `.claude/` and `CLAUDE.md`.55 Prefer a verified pinned checkout as the local marketplace. Never silently copy over56 `.claude/skills/harness`, `harness-evolve`, or `evolve`, including linked/shared roots.573. **Audit before generating.** Read [factory and evolution workflow](references/workflow.md).58 Inspect `.claude/agents/`, `.claude/skills/`, `CLAUDE.md`, and existing orchestration.59 Report new-build, extension, maintenance, or v1 migration scope. Reuse existing roles60 and skills instead of creating same-purpose duplicates. Preserve user edits and output language.614. **Choose a mode from the control flow and actual host capabilities.** Deterministic62 lists/loops use the upstream Workflow mode only if `Workflow` is exposed; repeated63 expert negotiation uses named agents plus messaging/tasks only if those tools are64 present; one-shot independent work uses available subagents. Check the actual tools,65 supported parameters, models, and budget before execution. Upstream documentation is66 not proof that this session supports `Workflow`, `Agent(name:)`, or `fable`. If a67 required capability is absent, report blocked; do not invent APIs or silently fall back.685. **Generate the smallest approved team.** Choose one of pipeline, fan-out/fan-in,69 expert pool, producer-reviewer, supervisor, or bounded hierarchical delegation.70 Define reusable roles under `.claude/agents/<name>.md`; methods and exactly one71 orchestrator under `.claude/skills/<name>/SKILL.md`. Specify inputs, outputs, schemas,72 ownership, failure/missing-result policy, and measurable completion. Keep `CLAUDE.md`73 to trigger pointers and change history. Read-only reviewers get no Write/Edit tools.746. **Keep execution and feedback bounded.** Start with a small sample and explicit token,75 iteration, and concurrency limits. Provider calls and generated commands retain host76 approval policies. Treat failed, missing, or `uncertain` outputs as unresolved, not77 passes. Freeze producer artifacts at phase boundaries before consumers use them.78 For feedback on an existing team, use plugin `/harness:evolve` (or direct-copy79 `/harness-evolve`), change one cause at a time, and record before/after evidence.80 Structural redesign returns to the factory; evolve does not silently add/remove roles.817. **Verify the generated result, not merely file presence.** Check frontmatter,82 cross-references, actual runtime tools, producer/consumer contracts, and trigger cases.83 Remove obsolete v1 constructs only after confirming the target runtime's replacement84 capabilities. Run a scoped task; compare with/without-skill results only with execution85 approval and report real artifacts, errors, timing/tokens when supplied, and unscored86 failures. A passed offline preflight does not prove plugin loading or team quality.8788## Examples8990### Inspect a pinned checkout and project without execution9192```bash93python3 .agent-skills/revfactory-harness/scripts/preflight.py --json \94 --repo /path/to/harness \95 --expect-commit 92d9f1b29e7af81acebf51a9fcd1832e739a1414 \96 --project /path/to/project97```9899### Build or evolve after the approved plugin install100101In the selected Claude Code session, request `revfactory/harness로 문헌 검토 팀을 구성해줘`.102After a real run, invoke `/harness:evolve` with the failed output and requested correction.103Keep the initial team, observed delta, adopted change, and the new verification result.104105## Best practices106107- Progressive disclosure: keep this front door small; load only the relevant pinned108 upstream reference linked from the local workflow guide.109- Select available models by task difficulty and cost; never assign every role the110 most expensive tier or assume an upstream alias resolves on another host.111- Do not enable `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` as an automatic migration fix.112- Preserve `_workspace/` evidence and run IDs; do not recreate missing receipts.113- Upstream's v1 self-reported A/B gains are not evidence for this v2 integration.114115## References116117- [Pinned provenance, installation modes, and host boundaries](references/upstream.md)118- [Factory, migration, quality, and evolution workflow](references/workflow.md)119- [Audited public-file SHA-256 manifest](references/source.json)120- [Behavioral routing and approval cases](evals/evals.json)