smithery/neversight

decision-envelope

Using the decision envelope pattern for structured thinking.

Installation

$ npx skills add smithery/neversight --skill decision-envelope

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from smithery/neversight · top by installs.

npx skills add smithery/neversight

Browse all from smithery/neversight

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 3,113 B
  • docs SUMMARY.md 85 B

History

  1. First recorded snapshot · 0 installs

SKILL.md

Skill: TraceMem Decision Envelopes

Purpose

This skill teaches how to correctly manage the lifecycle of a Decision Envelope. The Decision Envelope is the mandatory boundary for all governed operations in TraceMem.

When to Use

  • Every time you intend to perform a task that requires reading private data or affecting system state.
  • At the very beginning of a new task or workflow.
  • When an existing decision has been closed and you need to perform follow-up actions (start a new decision).

When NOT to Use

  • Do not nest decisions (e.g., do not open a decision if you are already inside one, unless explicitly starting a sub-task that requires independent audit).
  • Do not open a decision for purely computational tasks (e.g., formatting text) that require no external data or side effects.

Core Rules

  • One Decision, One Lifecycle: A decision must be explicitly created (open), operated upon, and then explicitly closed (commit or rollback).
  • Mandatory Intents: You must provide a structured intent string (e.g., customer.onboarding.verification) that describes why this decision exists.
  • Mandatory Automation Mode: You must certify the automation_mode (e.g., propose, approve, autonomous).
  • Close Required: A decision left open is a "zombie" decision. You must ensure decision_close is called in finally blocks or error handlers.

Correct Usage Pattern

  1. Create the Envelope:

Call decisioncreate with: - intent: A dot-separated string (e.g., financial.report.generate). - automationmode: One of propose, approve, override, autonomous. - actor: Your agent identity.

Result: You receive a decision_id.

  1. Operate within the Envelope:

Perform all decisionread, decisionevaluate, and decisionwrite calls using the decisionid.

  1. Close the Envelope:

- If successful: Call decisionclose with action: "commit". - If error/aborted: Call decisionclose with action: "rollback" (or abort if strictly tracking abandonment).

Note: Writes are only permanently applied when the decision is committed (depending on system configuration, but logically, the decision is not "done" until committed).

Common Mistakes

  • Forgetting to close: Leaving decisions open forever consumes resources and confuses auditors.
  • Vague Intents: Using generic intents like task.do or agent.act. Use specific, domain-relevant intents (user.password.reset, invoice.payment.process).
  • Wrong Automation Mode: Claiming autonomous when the task requires human oversight, or propose when you intend to execute immediately.

Safety Notes

  • Audited Lifecycle: The timestamps of open and close are recorded. Long-running open decisions may trigger alerts.
  • Fail Closed: If your process crashes, the decision remains open (and potentially locks resources). Always wrap your workflow in a try/finally block to ensure decision_close is attempted.