memory-query-capture
Capture query synthesis as durable pages in semantic memory
Best use case
memory-query-capture is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
It is a strong fit for teams already working in Codex.
Capture query synthesis as durable pages in semantic memory
Teams using memory-query-capture should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/memory-query-capture/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How memory-query-capture Compares
| Feature / Agent | memory-query-capture | Standard Approach |
|---|---|---|
| Platform Support | Codex | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Capture query synthesis as durable pages in semantic memory
Which AI agents support this skill?
This skill is designed for Codex.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
Related Guides
Cursor vs Codex for AI Workflows
Compare Cursor and Codex for AI coding workflows, repository assistance, debugging, refactoring, and reusable developer skills.
AI Agents for Coding
Browse AI agent skills for coding, debugging, testing, refactoring, code review, and developer workflows across Claude, Cursor, and Codex.
AI Agents for Marketing
Discover AI agents for marketing workflows, from SEO and content production to campaign research, outreach, and analytics.
SKILL.md Source
# memory-query-capture Capture valuable assistant synthesis from conversation into durable semantic memory pages. This is the compounding primitive — explorations accumulate as reusable knowledge rather than evaporating into chat history. ## When to Use Any time an assistant response produces analytical output worth preserving: comparisons, gap analyses, architectural assessments, synthesis narratives. Invoked explicitly by the user, suggested ambiently by the system, or chained automatically by other skills. ## Parameters ### --consumer (optional) Consumer ID to write against. Resolved via ADR-021 D4 precedence: explicit > wrapper > auto-detect. ### --type (optional) Page type: `synthesis`, `comparison`, `analysis`, or `gap`. If omitted, determined by heuristic or user prompt. ### --title (optional) Page title. If omitted, derived from content via headline extraction or summarization. ## Invocation Modes ### Ambient mode Suggested automatically when a query response exceeds a length/value threshold. The system presents a capture prompt; the user confirms or dismisses. ### Explicit mode User invokes directly: "save this as a page", "capture that comparison", or calls `memory-query-capture` with parameters. ### Skill-chain mode Any skill can call `memory-query-capture` as a final step after producing analytical output. The calling skill passes `--type` and `--title` to skip heuristics. ## Operation 1. **Capture synthesis** — extract the last assistant synthesis from conversation state. 2. **Determine page type** — from `--type` flag, content heuristic, or user prompt. 3. **Resolve consumer** — determine target consumer via ADR-021 D4 precedence (explicit > wrapper > auto-detect). 4. **Resolve destination** — read `memory.topology.derivedPages` from consumer's `manifest.json` and select the matching subdirectory (`derivedPages.synthesis`, `derivedPages.comparison`, etc.). 5. **Title the page** — from `--title` flag or derived from content (first heading, key noun phrase, or summarized label). 6. **Add cross-references** — identify source pages cited in the synthesis and insert cross-reference links per the consumer's `crossRefStyle` setting. 7. **Write page** — create the Markdown file at the resolved destination path. 8. **Update index and log** — call `memory-log-append` with `--op query-capture` and metadata about the captured page. 9. **Mark provenance** (optional) — record query origin, model actor, and sources cited in a provenance block at the end of the page. ## Page-Type Heuristics When `--type` is not provided, the skill inspects the synthesis content: | Content pattern | Assigned type | |-----------------|---------------| | Comparison table, side-by-side columns | `comparison` | | Narrative prose with connective reasoning | `synthesis` | | Gap analysis, missing coverage, "what's absent" | `gap` | | Bullet-point assessment, evaluation criteria | `analysis` | If the heuristic is ambiguous, prompt the user to select. ## Consumer Resolution Follows ADR-021 D4 precedence: 1. **Explicit** — `--consumer` flag provided on invocation. 2. **Wrapper** — a parent skill passes consumer context down the chain. 3. **Auto-detect** — inspect `.aiwg/frameworks/registry.json` for the active consumer with a `memory.topology` contract. If no consumer can be resolved, abort with a clear error requesting `--consumer`. ## Cross-Reference Style The consumer's `manifest.json` declares a `crossRefStyle` — one of: - `wikilink` — `[[page-name]]` - `markdown` — `[page name](relative/path.md)` - `at-mention` — `@consumer/path/to/page.md` The skill reads this setting and formats all source-page links accordingly. ## Error Handling - If no synthesis content is available in conversation state, report "nothing to capture" and exit. - If the destination directory does not exist, create it before writing. - Log write failures via `memory-log-append` are **non-blocking** — the page write is the primary operation. - If consumer topology lacks a `derivedPages` mapping for the resolved type, fall back to the consumer's base memory directory. ## Schema Reference @semantic-memory/schemas/memory-log-event.md ## Examples ``` # Explicit capture after a deep comparison memory-query-capture --consumer research-complete --type comparison --title "Redis vs Valkey feature matrix" # Ambient capture (system-suggested, user confirms) > Your comparison of caching strategies looks worth preserving. Capture as a page? [Y/n] memory-query-capture --type synthesis # Skill-chain: architecture-evolution calls capture after producing assessment memory-query-capture --consumer sdlc-complete --type analysis --title "Microservices migration readiness" ```
Related Skills
rlm-query
Spawn sub-agent to process focused context and return structured result
research-query
Search the local research corpus, read matching findings, and synthesize an answer with inline citations to REF-XXX sources. The "query" operation for the research pipeline.
ralph-memory
Manage Al semantic memory entries — list, query, and clear lessons learned across loop iterations
provenance-query
Query provenance chains to trace artifact derivation and impact
memory-log-render
Generate a human-readable Markdown view from a consumer's JSON Lines event log
memory-log-append
Append a structured event to a consumer's semantic memory log
memory-ingest
Ingest a source into any consumer's semantic memory by reading the topology contract
memory-forensics
Volatility 3 memory forensics workflows covering acquisition with LiME and WinPmem, and structured analysis using Volatility 3 plugin reference
debug-memory
Query and manage the executable feedback debug memory
aiwg-orchestrate
Route structured artifact work to AIWG workflows via MCP with zero parent context cost
venv-manager
Create, manage, and validate Python virtual environments. Use for project isolation and dependency management.
pytest-runner
Execute Python tests with pytest, supporting fixtures, markers, coverage, and parallel execution. Use for Python test automation.