Output Critic Protocol

Evaluate the output by type, score each criterion, make an accept/reject decision, and suggest concrete improvements. Goal: prevent weak output from reaching the next step.

Workflow

1. Detect output type
2. Apply type-specific criteria
3. Score each criterion
4. Calculate overall score
5. Make accept / conditional / reject decision
6. Suggest improvements

Acceptance Threshold

Type-Specific Criteria

Code

Report / Written Content

Plan / Task List

Data / Table

Output Format

OUTPUT CRITIC
Type     : [output type]
Decision : ACCEPT / CONDITIONAL / REJECT
Score    : [X/10]

## Criterion Scores

| Criterion | Score | Note |
|-----------|-------|------|
| [Criterion 1] | X/10 | [short note] |
| [Criterion 2] | X/10 | [short note] |
| **Overall** | **X/10** | |

## Strengths
- [What was done well — specific]

## Weaknesses
- [What is missing / wrong — specific]

## Improvement Suggestions
1. [Concrete action — what to do, where]
2. [Concrete action]

## Next Step
[Accept -> proceed | Conditional -> fix X | Reject -> apply suggestions, resubmit]

Re-Evaluation

When improved output is resubmitted:

RE-EVALUATION
Previous score: X/10
New score     : Y/10
Change        : +N points
Improved      : [which criteria]
Still open    : [remaining issues if any]

When to Skip

• User said "quick and dirty, doesn't need to be perfect"
• Prototype / draft stage (user explicitly stated)
• Single-line simple output

Guardrails

• Never accept security issues — hardcoded secrets = automatic REJECT regardless of other scores.
• Be specific in suggestions — "improve code" is useless; "move API key to env var at line 12" is actionable.
• Cross-skill: works with task-decomposer (validates plan quality), output-critic is the quality gate before task completion.

References

• references/EXAMPLES.md — scored critiques of code, reports, plans, and data outputs. Read it when the expected output format or level of detail is unclear.

🔗 Next Steps & Handoffs

Ecosystem: @ecosystem-orchestration — Agent Orchestration.

Workflows:

• Complex Task Execution Flow (orch-complex-task, step 5 of 6): next → @agent-reviewer (optional).
• Decision Quality Flow (reason-decision, step 5 of 5, via @ecosystem-reasoning): last step → once it passes, complete the workflow and report the outcome.

Handoff contract: pass results to the next skill through an inline Handoff block in your reply with the fields skill, workflow, created_at, inputs, summary, and next (the handoff contract of @ecosystem-orchestration). If a next skill is not installed, continue with its step from the ecosystem map, or install it with npx skills add fatih-developer/fth-skills --skill <name> after the user agrees.