gratitud-post-proceso

name	gratitud-post-proceso
version	0.2.0
description	Redacta agradecimientos post-proceso diferenciados por persona, evidencia de interaccion, voz de marca y lint offline contra FOMO, hustle, servilismo y promesas no verificables.
owner	JM Labs
triggers	["gratitud-post-proceso","agradecimiento","thank-you"]
allowed-tools	["Read","Write","Edit","Bash"]

Gratitud Post Proceso

Purpose

Use this skill after an interview, selection process, workshop, review panel, or professional conversation when the user needs a thank-you note that is specific, grounded in supplied evidence, and safe to send. The skill produces differentiated messages per recipient, preserves factual boundaries, and validates tone with scripts/lint_gratitud.py.

Inputs Expected

• Process or meeting context.
• Recipient name, role, and relationship to the process.
• Evidence from the interaction: topics discussed, contribution, decision stage, or next step.
• Channel constraints: email, LinkedIn, WhatsApp, note, or internal message.
• Brand voice constraints and any promises that must not be made.

Outputs Expected

• One message per recipient or persona.
• Recipient-specific evidence line.
• Subject or opening line when the channel needs it.
• Tone and risk notes for FOMO, hustle, servility, overpromising, or invented process details.
• Validation command evidence when a JSON packet is provided.

Procedure

Discover

Identify the process stage, channel, recipient, relationship, evidence from the interaction, and whether the user wants a message draft or a reusable template.

Analyze

Apply assets/recipient-differentiation-policy.json, assets/evidence-policy.json, assets/brand-voice-policy.json, and assets/promise-boundary-policy.json before drafting.

Execute

Write concise gratitude that names a real contribution, avoids pressure, and states next steps only when the user supplied them. Do not invent interview details, relationship warmth, availability, outcomes, or deadlines.

Validate

Run the deterministic fixture suite:

bash skills/gratitud-post-proceso/scripts/check.sh

For a specific packet:

python3 skills/gratitud-post-proceso/scripts/lint_gratitud.py --input <packet.json>

Assets

• assets/recipient-differentiation-policy.json
• assets/evidence-policy.json
• assets/brand-voice-policy.json
• assets/promise-boundary-policy.json
• assets/output-contract.json

Quality Criteria

• Each message has a named recipient or explicit recipient archetype.
• Each message includes at least one interaction-specific evidence detail.
• Tone is grateful, concrete, and calm; it avoids servility, pressure, and performative urgency.
• Brand phrases are not stacked; at most one distinctive phrase is allowed per message.
• Follow-up or availability statements are conditional on user-provided facts.
• Missing process evidence produces a partial or blocked output instead of invented detail.

Edge Cases

• No recipient: produce a blocked checklist and ask for the target person or persona.
• No interaction evidence: provide a generic-safe template and mark missing evidence.
• Multiple recipients: differentiate by role and contribution, not by flattery intensity.
• User asks to pressure the recipient: remove FOMO and state a neutral follow-up.
• User asks to apologize excessively: rewrite to gratitude without servility.
• User asks to imply acceptance, offer, or commitment: block unless supplied as fact.

Assumptions and Limits

• The skill improves professional communication; it does not guarantee process outcomes.
• If evidence is unavailable, mark the message as template-only or ask for context.
• Do not include private contact details in examples or fixtures.

Scripts

scripts/lint_gratitud.py --input <json> validates gratitude packets for recipient specificity, interaction evidence, tone blockers, promise boundaries, and output sections. It also accepts plain text for ad hoc tone linting. scripts/check.sh runs deterministic valid and invalid fixtures offline.

Related Skills

• proceso-seleccion-orchestrator
• negociacion-oferta

Evidence Requirements

• Tie every specific thanks, topic, next step, or process claim to user-provided evidence.
• Mark missing evidence as a blocker or assumption.
• Report validation commands and results when a machine-readable packet is used.

Update-Safety Notes

• Default to draft text unless the user explicitly asks to edit files.
• Preserve the user's facts, names, and relationships.
• Do not add network checks, wall-clock timestamps, random IDs, or live process-status claims to validation.