borghei/Claude-Skills / engineering/agenthub
agenthub
VerifiedRuns a multi-agent job as a dependency graph: validates the DAG, tracks each agent's state on a board, then ranks and merges their outputs. Use when spawning parallel agents, monitoring a run, or merging results.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.8 KB |
| assets/sample_session.json | config | 4.5 KB |
| assets/sample_workflow.json | config | 1.5 KB |
| references/multi-agent-patterns.md | markdown | 5.6 KB |
| references/operations-and-quality.md | markdown | 2.9 KB |
| references/orchestration-core.md | markdown | 4.9 KB |
| scripts/board_manager.py | script | 10.9 KB |
| scripts/dag_analyzer.py | script | 13.5 KB |
| scripts/result_ranker.py | script | 11.0 KB |
| scripts/session_manager.py | script | 12.7 KB |
| skills/board.md | markdown | 2.2 KB |
| skills/eval.md | markdown | 2.2 KB |
| skills/init.md | markdown | 2.0 KB |
| skills/merge.md | markdown | 2.3 KB |
| skills/run.md | markdown | 2.1 KB |
| skills/spawn.md | markdown | 1.9 KB |
| skills/status.md | markdown | 2.1 KB |
skills/eval.md
View on GitHub# Sub-Skill: Evaluate Agent Output
**Parent:** agenthub
**Trigger:** "evaluate output", "check agent quality", "grade results"
## Purpose
Evaluate the quality of an agent's output against defined criteria. Quality gates prevent low-quality outputs from propagating to downstream agents.
## Workflow
### Step 1: Define Evaluation Criteria
Each agent can have evaluation criteria in its config:
```json
{
"eval_criteria": {
"completeness": "All required output fields populated",
"accuracy": "Claims supported by evidence",
"format": "Output matches expected JSON schema",
"relevance": "Output addresses the assigned task"
},
"quality_threshold": 0.7
}
```
### Step 2: Run Evaluation
Score the output on each criterion (0.0 to 1.0):
| Criterion | Check | Score |
|-----------|-------|-------|
| Completeness | All declared outputs present and non-empty | 0.0 - 1.0 |
| Format compliance | Output matches expected schema | Pass/Fail |
| Length adequacy | Output has sufficient depth (not trivially short) | 0.0 - 1.0 |
| Relevance | Output content relates to the task description | 0.0 - 1.0 |
| Consistency | Output does not contradict inputs | 0.0 - 1.0 |
Composite score = weighted average of criteria scores.
### Step 3: Gate Decision
| Composite Score | Decision |
|----------------|----------|
| >= threshold | PASS -- output flows to downstream agents |
| >= threshold - 0.15 | REVISE -- retry with feedback on what to improve |
| < threshold - 0.15 | FAIL -- agent marked as failed |
### Step 4: Revision Feedback (If REVISE)
Generate specific feedback for the agent retry:
- Which criteria scored low
- What was missing or incorrect
- Concrete improvement instructions
### Step 5: Log Evaluation
Record the evaluation result in the session for audit:
- Per-criterion scores
- Composite score
- Decision (PASS/REVISE/FAIL)
- Feedback provided (if any)
## Inputs
| Input | Required | Description |
|-------|----------|-------------|
| Agent output | Yes | The output to evaluate |
| Eval criteria | Yes | Criteria from agent config |
| Quality threshold | No | Override default threshold |
## Outputs
- Per-criterion scores
- Composite quality score
- Gate decision (PASS / REVISE / FAIL)
- Revision feedback (if applicable)
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: 472k | Updated | Claude CodeCodex | Unverified | 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 |