evaluate-skill

A command for reviewing a recent run of a coding-agent skill and recording a human rating, problems, and suggestions.

In plain words
What is it for?
Use it to review the latest run, all recent runs, or a run from a specific date, then save feedback to the run log.
Why use it?
It makes it easier to capture what worked or went wrong while the details of a skill run are still available.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/athola/claude-night-market/evaluate-skill
Clone the repo
git clone --depth 1 https://github.com/athola/claude-night-market
Per session 16 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 941 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00016 $0.00941
Opus 5 $0.00008 $0.00470
Sonnet 5 $0.00003 $0.00188
Haiku 4.5 $0.00002 $0.00094

Measured yesterday against content hash e8a253aefcf6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evaluate-skill scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/abstract/commands/evaluate-skill.md · 148 lines

How it starts

The opening of the file, as written. The whole thing — 148 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Evaluate Skill Execution

Manually evaluate a recent skill execution for qualitative feedback.

Usage

/abstract:evaluate-skill <skill-name>              # Evaluate most recent execution
/abstract:evaluate-skill <skill-name> --all        # Evaluate all recent executions
/abstract:evaluate-skill <skill-name> --date YYYY-MM-DD  # Evaluate specific date

Purpose

Captures human-in-loop qualitative feedback about skill effectiveness to improve the self-improvement loop. This data feeds into:

  • Log aggregation (Phase 3)
  • LEARNINGS.md pattern detection
  • /improve-skills prioritization (Phase 5)

What It Does

  1. Loads recent executions from ~/.claude/skills/logs/<plugin>/<skill>/
  2. Presents execution summary (timestamp, duration, outcome, context preview)
  3. Prompts for evaluation:
    • Effectiveness rating (1-5)
    • Friction points encountered
    • Improvement suggestions
  4. Updates log entry with evaluation metadata
  5. Tracks evaluation completion (prevents duplicate evaluations)

Evaluation Questions

1. Effectiveness Rating (1-5)

  • 5 - Exceptional: Skill exceeded expectations, no friction
  • 4 - Effective: Skill achieved goal with minor friction
  • 3 - Adequate: Skill worked but had noticeable inefficiencies
  • 2 - Poor: Skill partially failed or required significant workarounds
  • 1 - Failed: Skill did not achieve intended outcome

2. Friction Points

What slowed you down or caused issues?

  • Skill missing key information?
  • Instructions unclear or incomplete?
  • Wrong tool recommendations?
  • Excessive token usage?
  • Outdated information?
  • Missing error handling?

3. Improvement Suggestions

What would make this skill better?

  • Add examples?
  • Simplify language?
  • Update workflow steps?
  • Add error recovery guidance?
  • Split into smaller skills?
  • Add context-specific variations?

Output

Updates the log entry with:

{
  "qualitative_evaluation": {
    "evaluated_at": "2026-01-08T04:00:00Z",
    "rating": 4,
    "friction_points": ["Missing error handling for X", "Y was unclear"],
    "improvement_suggestions": ["Add example for Z", "Simplify step 3"],
    "evaluator_notes": "Worked well but needed clarification on..."
  }
}

Read the full file on GitHub · 148 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 148 lines · 16 tokens per session scan A e8a253aefcf6

Subscribe to this mod's changes

evaluate-skill is a command published in the GitHub repository athola/claude-night-market (332 stars, last pushed 5d ago), licensed MIT. It adds 16 tokens to every session and 941 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.