agentleFS
marcopavezmarcopavez/claude-impact-lab/.claude/skills/mentor-checklist/SKILL.md

mentor-checklist

Skill0 starsChanged 7 months agoMITView on GitHub

Use before tagging any milestone (v0.1-mvp-call, v0.5-solid, v1.0-demo-final) and any time Marco asks "cómo estamos contra los sub-checks de mentor". Runs the 10 mentor sub-checks (A1-A6 in M1+M2, B1-B4 in M3+M4) over the current repo state and reports green/yellow/red per sub-check with concrete evidence paths and minimum gap-closer. Mirrors logic of the rubrica-auditor agent but scoped strictly to mentor (40% peso). Triggers - "mentor checklist", "estamos listos para v0.5", "cómo vamos en mentor".

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No marques 🟢 por intención.

    line 85
  2. Distinctive:

    No mezcles peso mentor con juez en el score.

    line 86
  3. Very specific:

    No mezcles este check con rubrica-auditor (mentor + juez).

    line 87

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill — mentor-checklist (10 sub-checks · 40% peso final)

Cuándo invocar

Antes de cualquier tag de hito o cuando Marco pregunta el estado contra mentor. No invocar para juez (J1-J3) — esa es otra skill / agente (rubrica-auditor cubre ambos).

Por qué solo mentor

Mentor define quiénes son los 12 finalistas (top 4 por vertical sobre score_mentor). El primer desempate es M3 (35% peso). Si mentor no pasa, juez no aplica. Por eso mentor merece su propia skill afilada.

Sub-checks (mapa breve, detalle en docs/PLAN.md §12)

IDPertenecePeso internoQué se mide
A1M1 (20%)sin jergapitch / readme legible para no-tech
A2M1segmento explícito"adultos mayores 65+ Chile, 2.4M INE" literal
A3M1canal adopcióncall forwarding GSM + PWA cuidador
A4M1impacto cuantificablemétrica medible (tiempo detección 72h → real)
A5M2 (20%)≥2 fuentes oficialesBCN + CMF + Sernac + CSIRT en prompts/RAG
A6M2sin alucinacionestool_choice required + citations[] minItems 1 + post-validator
B1M3 (35%)system prompts dedicados≥6 archivos en prompts/ con front-matter
B2M3≥2 tools/MCPsapps/mcp-* + tools registrados
B3M3≥3 mensajes consola Anthropic en ventanascreenshots en docs/EVIDENCIA/
B4M4 (25%)demo end-to-endapps/server corriendo + 3 llamadas pre-validadas + video backup

Procedimiento

  1. Leé docs/EVENT/RUBRICA.md (autoridad), docs/PLAN.md §12, CLAUDE.md "Defensas".
  2. Buscá evidencia por sub-check:
    • A1: grep en README.md y pitch deck de términos crudos sin glosar (vishing, TTL, DID, MCP).
    • A2: grep literal "adultos mayores" en docs/PROYECTO.md + número INE.
    • A3: grep **21* en docs + screenshot PWA.
    • A4: número medible en pitch.
    • A5: contar fuentes únicas citadas en prompts/*.md y apps/eval/golden-set/*.jsonl.
    • A6: grep tool_choice.*required en codigo + schema con minItems: 1 + existencia de citation-validator agent.
    • B1: count files en prompts/ con frontmatter válido (---\nname:...).
    • B2: list apps/mcp-*/ + grep tool definitions.
    • B3: existencia de docs/EVIDENCIA/anthropic-console-*.png con timestamp ≥2026-05-06.
    • B4: git log --since=2026-05-06 muestra commits de app + existe video en docs/EVIDENCIA/demo-*.mp4.
  3. Clasificá:
    • 🟢 green: evidencia presente, verificable.
    • 🟡 yellow: parcial; nombrá la pieza faltante.
    • 🔴 red: sin evidencia.
  4. Output: tabla + score estimado + top-3 gaps por ROI.

Output

## Mentor Checklist · {fecha}

| Sub-check | Estado | Evidencia | Gap mínimo |
|---|---|---|---|
| A1 sin jerga | 🟢 | README L20 glosario completo | — |
| A2 segmento | 🟢 | docs/PROYECTO.md §"Ciudadano" | — |
| A3 canal | 🟡 | call forwarding documentado | falta screenshot PWA installable |
| A4 impacto | 🟢 | "72h → tiempo real durante llamada" + 2.4M INE | — |
| A5 ≥2 fuentes | 🟢 | 7 fuentes únicas en prompts/regulatory.md | — |
| A6 sin alucinaciones | 🟡 | tool_choice + citations[] OK | falta corrida automática del validator en CI |
| B1 ≥6 prompts | 🟢 | 7 archivos en prompts/ con frontmatter | — |
| B2 ≥2 tools/MCPs | 🟢 | mcp-wiki-legal + mcp-cmf + 8 tools SDK | — |
| B3 ≥3 mensajes consola en ventana | 🔴 | sin screenshots en docs/EVIDENCIA/ | tomar 3 screenshots con timestamps post 2026-05-06 |
| B4 demo end-to-end | 🟡 | server corre, faltan 3 llamadas validadas + video backup | grabar 3 calls + video 90s |

**Score estimado mentor:** 78 / 100
- M1: 95% (A3 corrige fácil)
- M2: 90% (A6 yellow se cierra con CI)
- M3: 80% (B3 está rojo — bloqueante)
- M4: 60% (B4 yellow — mayor inversión)

**Top-3 gaps por ROI:**
1. 🔴 B3 — abrir consola Anthropic, pegar 3 prompts reales, screenshot timestamped. **15 min.**
2. 🟡 B4 — grabar las 3 llamadas pre-validadas como backup. **45 min.**
3. 🟡 A3 — capturar PWA en home screen Android. **5 min.**

**Riesgo descalificación:** ninguno (Claude motor verificable, código en ventana).

Anti-patrones

  • No marques 🟢 por intención. Solo por archivo/grep que existe.
  • No mezcles peso mentor con juez en el score.
  • No mezcles este check con rubrica-auditor (mentor + juez). Esta skill es scoped.
  • Si docs/EVENT/RUBRICA.md no existe, parás y avisás.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.