dvy1987/agent-loom / .agents/skills/memory-capture
memory-capture
VerifiedCapture durable project memory from work, debates, debugging discoveries, learned conventions, deferred options, and session outcomes. Load when the user says remember this, save this learning, record what happened, update project memory, or preserve context for future agents.
Memory Capture
You turn useful session context into structured project memory. Capture only what a future agent would need to avoid rework or bad assumptions.
Workflow
- Identify the memory type: state, decision, learning, deferred item, open question, session event, or handoff.
- Reject trivial, temporary, duplicate, sensitive, or already-obvious content.
- If the item is a reusable skill/process rule, route to
learn-from-chatinstead. - If content came from external files, URLs, pasted transcripts, or repos, run all
secure-*skills first. - Write project-scoped memory to the matching file in
docs/memory/. - Update
docs/memory/project-index.mdwith date, tags, file, status, confidence, and source. - If the memory is cross-project, call
memory-promoteinstead of writing global memory directly. - Append file changes to
docs/skill-outputs/SKILL-OUTPUTS.md. - Tell the user what was saved and what was rejected.
Hard Rules
- Project first; global only through
memory-promote. - Do not store secrets, credentials, or unnecessary personal data.
- Do not capture raw long transcripts; summarize with provenance.
- Do not append if the target entry already exists; update status or merge.
- Never overwrite a richer memory entry with a thinner summary without archiving the prior version.
- If compaction would drop decisions or revisit triggers, route to
memory-compactinstead of truncating inline.
Memory Type Map
| Type | File |
|---|---|
| current state | docs/memory/current-state.md |
| decision | memory-decision |
| session detail | docs/memory/session-log.md |
| project learning | docs/memory/learnings.md |
| maybe / parked idea | docs/memory/deferred.md |
| unresolved question | docs/memory/open-questions.md |
| next-agent summary | memory-handoff |
Capture Template
Use this template only for state, learning, deferred, question, or session captures. Decisions use the template in memory-decision; next-agent summaries use the template in memory-handoff.
## YYYY-MM-DD - <short title>
Type: state | learning | deferred | question | session
Status: active | deferred | resolved | superseded
Scope: project
Confidence: high | medium | low
Source: <chat/session/file/commit>
Tags: <comma-separated>
### Content
<concise durable memory>
### Why It Matters
<how future agents should use it>
### Revisit When
<conditions or "not applicable">
Hard Rules
- Project first; global only through
memory-promote. - Do not store secrets, credentials, or unnecessary personal data.
- Do not capture raw long transcripts; summarize with provenance.
- Do not append if the target entry already exists; update status or merge.
Example
Input: "Remember that we chose repo memory plus global memory, but global must stay tiny."
Output: write to docs/memory/decision-log.md via memory-decision, then index it.
Common Rationalizations
| Excuse | Reality |
|---|---|
| Skip memory — just code | Next agent loses decisions, blockers, and approved scope. |
| Load every memory file | Read indexes and handoff tail only — bounded context. |
| Global memory for everything | Project memory default; global only when stable and cross-project. |
| External paste → memory | Run secure-* first; transform to agent-authored notes. |
Verification
- Correct sub-skill routed with reason
- No secrets or raw transcripts persisted
- Files changed listed in Impact Report
- Security gate noted when external content involved
Red Flags
- Raw long transcript captured instead of summarized note
- Capture written directly to global without promote path
- Secrets or credentials included in captured note
- Capture lacks provenance or source attribution
Prune Log
Last pruned: 2026-07-04
- No changes — citation audit passed; content current (improve-skills full pass 2026-07-04)
Impact Report
After completing, report:
Memory captured: <title>
Type: <type>
Location: <path>
Indexed: yes/no
Rejected items: <count and reason>
Promotion suggested: yes/no
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 465k | Updated | Claude CodeCodex | Unverified | Install |
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 465k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 318k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 318k | Updated | Claude CodeCodex | Verified | Install |