SpillwaveSolutions/research-knowledge-capture

Latest recorded source activity: September 12, 2026 at 17:22

SkillsMP catalog refreshed: November 12, 2026 at 12:52

skills collected: 9

GitHub stars: 2
GitHub forks: 1

Install with an AI assistant

Copy this prompt into the AI assistant you're using.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/SpillwaveSolutions/research-knowledge-capture.

Repository README

Repository README source: README.md

Read the source README

Expand full README

# Research Knowledge Capture (RKC) Layer 0 ContentPack for a research second brain. OKF Markdown + YAML is the source of truth. Agent Brain is a disposable index — see [`research-graph`](https://github.com/SpillwaveSolutions/research-graph). **Version:** 0.2.13. `rkc_ingest.py --set-resources FILE --source-hash H` gives a looked-up URL list to a source whose dump cites works by name only. See the [PRD](docs/prd/research-knowledge-capture-PRD.md). ## What it owns Eight nouns: `ResearchArea`, `Subject`, `ResearchTask`, `SourceDocument`, `ResearchQuestion`, `Claim`, `Evidence`, `Finding`. Registered rels: `has_subject`, `related_to`, `has_task`, `ingested_from`, `asks`, `answers`, `produced`, `asserts`, `evidenced_by`, `contradicts`, `supersedes`, `same_as`. Article bridge (`Article → draws_from → Finding`) is owned by **content-media**, not this pack. ## Install ``` claude plugin marketplace add SpillwaveSolutions/research-knowledge-capture claude plugin install research-knowledge-capture@rkc-plugin-marketplace ``` Hosts: Claude Code, Grok Build, Codex, Cursor, Agent Plugins 1.0. See [docs/HOSTS.md](docs/HOSTS.md). ## Commands | Host | Ingest | Extract | Pack | Retrieve | Search | Validate | Spine | | --- | --- | --- | --- | --- | --- | --- | --- | | Claude / Grok / Cursor | `/research-ingest` | `/research-extract` | `/research-pack` | `/research-retrieve` | `/research-search` | `/research-validate` | `/research-spine` | | Codex | `$research-ingest` | `$research-extract` | `$research-pack` | `$research-retrieve` | `$research-search` | `$research-validate` | `$research-spine` | ``` python3 scripts/rkc_validate.py --root sample-knowledge python3 scripts/rkc_pack.py subject.loop-policy.01J8X000000000000000000001 --root sample-knowledge --summary python3 scripts/rkc_search.py "false civic alerts" --root sample-knowledge --limit 5 --json python3 tests/test_rkc.py python3 tests/test_extract.py python3 tests/test_plugin.py python3 tests/test_bulk_fixes.py python3 tests/test_spine.py python3 tests/test_search.py ``` Public samples are **Northstar / Lumenfield fiction**. Live dumps stay in private trees. ## Bulk ingest - Lookup is `research/catalogs/ingest-index.json` (not a full-tree scan). Rebuild with `--rebuild-index`. - Archive-only ingest still creates the Subject and writes `has_task`. Pass `--area <slug>` for a ResearchArea + `has_subject`. - `index.md` / `README.md` titles are `parent/filename`. `--extractor-version` defaults to the current extractor, so omitting `--extract` on a re-run is idempotent. - Repair a 0.2.0 tree with `python3 scripts/rkc_spine.py --knowledge knowledge --link-tasks` then `--area-map areas.json`. - Heuristic extract skips files over 200 KB. Use `--force-large` only when you mean it. - One process per Subject (or per knowledge tree). Do not `pkill -f` a multiprocessing ingest; kill by PID. - `vendor` is free text. Convention: `grok` `gemini` `claude` `deepseek` `chatgpt` `article` `perplexity` `unknown`. - `--source-kind` defaults to `deep_research`. ## Retrieval ladder `rkc_search` (`rg` → scan) → `research-retriever` / `rkc_pack --summary` → BM25/Chroma → Kuzu last (Layer 1). Q&A retrieval is a sub-agent: the parent gets a summary card only. `research-graph` may feed seeds later and is not required for L0. ## Actor `grok-bot/research-knowledge-capture`. Isolated session. Ready PR. No force-push. No write to main.

Loading the repository README…

Skills in this repository

Market Research Analysts & Marketing Specialists Software Developers Computer Occupations, All Other Office Clerks, General

5 occupation categories · 100% classified

skill

occupation

description

updated

cursor-research-knowledge-capture

Market Research Analysts & Marketing Specialists

Bind a Cursor agent to RKC. Capture research into OKF. Do not project the graph.

2026-09-12

grok-bot-research-knowledge-capture

Market Research Analysts & Marketing Specialists

Bind a Grok Bot agent to RKC. Isolation, identity, fail-closed writes.

2026-09-12

research-pack

Computer Occupations, All Other

Build an RKC ContextPack. Structural max_hops=2 plus Finding→Claim→Evidence spine. Question roots allow one inbound hop on answers.

2026-09-12

research-retrieve

Market Research Analysts & Marketing Specialists

Spawn research-retriever for RKC Q&A. Parent gets a summary card only. Do not pack inline.

2026-09-12

research-search

Market Research Analysts & Marketing Specialists

Thin git-native RKC search (rg → scan). AND terms, type filter, limit, json. No Chroma or Kuzu.

2026-09-12

research-ingest

Office Clerks, General

Archive AI deep-research dumps and articles into RKC source-assets. Hash, copy, write SourceDocument + ResearchTask shells. Optionally run the Phase 2 extractor. Idempotent on bytes plus prompt hash plus extractor version.

2026-08-25

research-spine

Software Developers

Create missing Subject and ResearchArea nodes and write has_task / has_subject on an existing RKC tree. Idempotent repair for 0.2.0 archive-only ingest.

2026-08-25

research-extract

Software Developers

Segment an archived research dump, extract claims/evidence/findings, merge on claim_key, write a required PR summary. Fail-closed. Never auto-supersede accepted or verified nodes.

2026-08-22

research-validate

Software Quality Assurance Analysts & Testers

Fail-closed RKC validator. Unknown type or rel is an error. Accepted Claims need evidenced_by. Verbatim quotes must match the archived span.

2026-08-22

Showing 9 of 9 collected skills.