Self-Cognitive Meta Skill

Overview

Provide a structured self-cognitive response that verifies reasoning, captures lessons, and produces persistable updates.

Modes

Choose the smallest set of modes that match the request:

• •Preflight Verification: before execution or decisions.
• •Postflight Retro: after work is done or after errors.
• •Skill Extraction: when the user wants persistence or repeatability.

Core Rules

• •Separate facts from assumptions.
• •State risks and how to validate quickly.
• •Calibrate confidence (low/medium/high) with evidence.
• •Never include secrets or sensitive data in any output.
• •Keep outputs concise; prefer bullets.
• •Always include the JSON artifact.
• •If the template grows, move detailed guidance to references/ and keep this file lean.

Iteration Loop (Revolve)

Use an iterative loop to strengthen outputs over time:

1. •Draft with the required template.
2. •Compare against the required sections and JSON artifact.
3. •Capture rationalizations or omissions.
4. •Update the skill update proposal accordingly.
5. •Re-run pressure scenarios until compliant.

Required Output Template

Goal

One sentence describing what is being verified or improved.

Current context

• •Bullet facts from the conversation.

Verification

Assumptions

• •...

Risks and failure modes

• •...

Checks

• •...

Confidence

• •Level: low | medium | high
• •Why: ...
• •To raise confidence:
  • •...

Lessons learned

Technical

• •...

Process

• •...

Prompting and coordination

• •...

Persistable updates

Memory summary

• •User preferences:
  • •...
• •Project facts:
  • •...
• •Open questions:
  • •...

Skill update proposal

If a target skill exists, propose minimal patch text. If no target skill exists, propose a new skill outline and frontmatter suggestions.

Artifacts JSON

{
  "self_cognitive_meta": {
    "goal": "",
    "context_facts": [],
    "verification": {
      "assumptions": [],
      "risks": [],
      "checks": [],
      "confidence": {
        "level": "low",
        "why": "",
        "to_raise": []
      }
    },
    "lessons_learned": {
      "technical": [],
      "process": [],
      "prompting_and_coordination": []
    },
    "memory_summary": {
      "user_preferences": [],
      "project_facts": [],
      "open_questions": []
    },
    "skill_updates": {
      "target_skill_name": null,
      "proposed_patch": null,
      "new_skill_drafts": []
    }
  }
}

Common Mistakes

• •Skipping the verification section to save time.
• •Omitting the JSON artifact.
• •Mixing assumptions with facts.
• •Declaring confidence without evidence.
• •Treating the first draft as final instead of iterating.

Rationalization Table

Red Flags — STOP and Restart

• •“Skip verification to move faster.”
• •“I’ll add lessons learned later.”
• •“This is too simple to structure.”
• •“No JSON needed.”

Violating the letter of the rules is violating the spirit of the rules.

Example (concise)

Goal

Validate whether the deployment plan is safe to execute today.

Current context

• •Deployment touches auth config and rate limits.

Verification

Assumptions

• •Rate limit change is backward compatible.

Risks and failure modes

• •Auth flow could reject valid tokens.

Checks

• •Dry-run staging deploy and login.

Confidence

• •Level: medium
• •Why: staging verified, no prod traffic tested
• •To raise confidence:
  • •Run canary on 5% traffic

Lessons learned

Technical

• •Staging coverage missed token edge case.

Process

• •Add token-matrix checklist to deploys.

Prompting and coordination

• •Ask for rollout window upfront.

Persistable updates

Memory summary

• •User preferences:
  • •Wants explicit confidence levels.
• •Project facts:
  • •Auth uses token rotation.
• •Open questions:
  • •Canary tooling availability

Skill update proposal

• •Draft a rollout-checklist skill if repeated.

Artifacts JSON

{
  "self_cognitive_meta": {
    "goal": "Validate whether the deployment plan is safe to execute today.",
    "context_facts": ["Deployment touches auth config and rate limits."],
    "verification": {
      "assumptions": ["Rate limit change is backward compatible."],
      "risks": ["Auth flow could reject valid tokens."],
      "checks": ["Dry-run staging deploy and login."],
      "confidence": {
        "level": "medium",
        "why": "staging verified, no prod traffic tested",
        "to_raise": ["Run canary on 5% traffic"]
      }
    },
    "lessons_learned": {
      "technical": ["Staging coverage missed token edge case."],
      "process": ["Add token-matrix checklist to deploys."],
      "prompting_and_coordination": ["Ask for rollout window upfront."]
    },
    "memory_summary": {
      "user_preferences": ["Wants explicit confidence levels."],
      "project_facts": ["Auth uses token rotation."],
      "open_questions": ["Canary tooling availability"]
    },
    "skill_updates": {
      "target_skill_name": null,
      "proposed_patch": null,
      "new_skill_drafts": ["rollout-checklist"]
    }
  }
}