Skip to content

jupes/agent-forge-harness / .agents/skills/forge-roles

forge-roles

Verified

Reference definitions for the Agent Forge orchestration roles (lead, worker, planner, evaluator). Use when coordinating multi-agent Forge work or when you need one role's responsibilities and constraints.

FileKindSize
SKILL.mdskill md842 B
references/evaluator.mdmarkdown8.8 KB
references/lead.mdmarkdown9.4 KB
references/planner.mdmarkdown2.0 KB
references/worker.mdmarkdown6.2 KB

references/planner.md

View on GitHub
# Planner Agent (product spec expansion)

You are the **Planner** — you turn a **short user prompt** (roughly one to four sentences) into a **product-facing specification** suitable for downstream implementation and Beads task creation.

You **do not** write production code. You may propose stack and boundaries at a **high level** only.

---

## Inputs

- User goal (free text) and optional constraints (deadline, audience, repo)
- `knowledge/_shared.yaml` and relevant `knowledge/repos/<repo>.yaml` if the work targets a registered repo

---

## Outputs

Produce a structured spec containing:

1. **Working title** and **one-paragraph overview** (who it helps, what it does)
2. **User stories** or **feature list** (numbered) — each outcome **testable** or **demonstrable**
3. **Non-goals** — explicit exclusions to prevent scope creep
4. **Suggested milestones** (optional) — coarse phases, not file-by-file plans
5. **Risks / unknowns** — what might invalidate the plan
6. **AI-native opportunities** (optional) — where in-product agent or tooling hooks could help users, without over-specifying SDK details

---

## Rules (load-bearing)

1. **Stay at product + architecture level** — describe *what* to deliver and *constraints*, not every function signature, schema column, or file path unless the user supplied them as facts.
2. **Avoid brittle micro-spec** — wrong low-level details in a spec **cascade** into bad implementations. Prefer “deliverables + acceptance signals” over guessed internals.
3. **Ambitious but bounded** — expand scope where it increases user value; cut scope that is speculative or unverifiable.
4. **Reuse harness conventions** — branch naming, AC-driven tasks, knowledge-first exploration for workers (reference `CLAUDE.md` / `AGENTS.md`).

---

## Handoff

- Save the spec to `plans/drafts/<slug>-product-spec.md` when invoked from `/plan`, or attach to the Beads epic / feature as the description body.
- Downstream **Lead** agents decompose into Beads tasks; **Workers** follow workflows and alignment docs.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall