# Proof Memory

> Use ControlKeel proof bundles, typed memory, workspace snapshots, transcript summaries, and resume packets before closing or resuming work. Activate this when you need durable evidence or historical context.

- Skill: `aryaminus/proof-memory` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add aryaminus/proof-memory`
- Raw SKILL.md: https://api.skillmd.com/api/skills/aryaminus/proof-memory/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: Apache-2.0
- Author: aryaminus (https://skillmd.com/u/aryaminus)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/aryaminus/proof-memory

---


# Proof and Memory Skill

Use this skill when you need the durable system-of-record view instead of only the live session state.

## Workflow

1. Call `ck_context`.
2. Review `proof_summary`, `memory_hits`, `workspace_context`, `context_reacquisition`, `instruction_hierarchy`, `recent_events`, and `resume_packet`.
3. Use the proof bundle to understand deploy readiness, regression evidence, open findings, and rollback expectations.
4. Use `ck_memory_search` when prior decisions, checkpoints, or findings need explicit retrieval instead of relying only on passive memory hits.
5. Use `ck_memory_record` to preserve new decisions or operator intent that future agents should recover explicitly.
6. Use `ck_memory_archive` to retire stale or superseded memories so retrieval quality does not decay.

## Observability evidence

When closing or resuming work, include local observability loop evidence when relevant: saved evals, benchmark drafts, materialized scenarios, generated benchmark history, and advisory promotion candidates. Prefer the read-only `ck_observability` MCP surface for summaries and record durable checkpoints with `ck_memory_record`; do not treat advisory promotion candidates as executed changes.

## Additional resources

- [Proof workflow](references/proof-workflow.md)