dvy1987/agent-loom / .agents/skills/create-agent-prompt
create-agent-prompt
VerifiedCreate focused role prompts for agents in multi-agent topologies. Load when agent-builder needs role prompts for agents, or when a user asks to "create an agent prompt", "write a role prompt", "define agent identity", "write an agent role", "prompt for this agent", "write instructions for this agent", "agent persona". Scope: agent role prompts only (v1). System prompts, task prompts, and skill invocation prompts are future TODOs.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.5 KB |
| references/examples.md | markdown | 2.1 KB |
references/examples.md
View on GitHub# Create Agent Prompt — Full Worked Examples
Skill: `create-agent-prompt` | Enriched from SKILL.md (improve-skills pass, SKIP_RESEARCH).
## Example 1 — Documented workflow
**Input:** Create a role prompt for a "Code Review Agent" in a parallel fan-out topology that receives code diffs and produces review comments.
**Output:**
```
ROLE: Code Review Agent — reviews code diffs for correctness, conventions, and test coverage.
RESPONSIBILITIES:
- Review each diff for bugs, logic errors, and convention violations
- Check that tests exist for new functionality
- Produce structured review comments with severity (critical/warning/nit)
- Does NOT fix code, refactor, or make changes — review only
SKILLS: [code-review-crsp]
TOOLS: [Read, Grep, Bash (for running tests)]
INPUT: Code diff (unified format) from orchestrator agent
OUTPUT: Structured review (list of findings with file, line, severity, comment) to merge agent
HANDOFF PROTOCOL:
- On success: Pass review to merge agent with verdict (approve/request-changes)
- On partial: Flag files that couldn't be reviewed (e.g., binary) to orchestrator
FAILURE BEHAVIOR:
- If diff is unparseable: report to orchestrator, do not guess
- Never approve without reviewing — if unsure, request-changes with explanation
```
## Example 2 — Step-by-step execution
**Input:** "Run `create-agent-prompt` on [concrete task]"
**Agent actions:**
1. Gather Context
2. Write Role Prompt
3. Validate
4. Return
**Impact Report shape:**
```
Agent prompt created for: [agent name]
Topology role: [role in topology]
Handoff: [to whom]
Failure behavior: defined
Ready to embed in: AGENTS.md / architecture spec
```
## Example 3 — Gotcha application
**Input:** Task hits a non-obvious edge case
**Apply:**
- Role prompts are NOT system prompts — they define behavior within a topology, not global identity.
- Avoid "be helpful" or "be thorough" — these are noise. Be specific about what the agent does.
- If the agent's boundary overlaps with another agent, the topology is wrong — flag it to agent-builder.
-
---
See `SKILL.md` for hard rules and verification checklist.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 471k | Updated | Claude CodeCodex | Verified | Install |
| 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: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| 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 available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |