How do I install the Pkc Retrieve skill?

Run npx skillmds@latest add spillwavesolutions/pkc-retrieve in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Pkc Retrieve safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Pkc Retrieve?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Pkc Retrieve free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Pkc Retrieve?

spillwavesolutions (@spillwavesolutions) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Pkc Retrieve by spillwavesolutions — Agent Skill

Name: Pkc Retrieve
Author: spillwavesolutions

Pkc Retrieve

Spawn the knowledge-retriever sub-agent to search and pack project memory without contaminating parent context. Use before answering from the OKF tree, or when starting work that needs Feature/Decision context.

File contents

PKC Retrieve

Parent-facing. The parent must not run pkc_search / pkc_pack itself for retrieval. Spawn the knowledge-retriever sub-agent (Claude Code Task/Agent tool, or host equivalent) and consume only the Retrieval card.

Do not pull raw packs into the parent

Do not run pkc_search.py or pkc_pack.py in the parent turn for retrieval.
Do not paste search hit lists, mermaid graphs, or ranked pack bodies into the parent working context.
Progressive disclosure (--tiny, hop caps) clips the graph; it does not isolate the parent. Isolation is the child agent.

Spawn

Pass the child:

Input	Required	Notes
query	one of query / seed	Free-text question or keywords
seed path	one of query / seed	Absolute in-bundle path, e.g. `/features/user-authentication.md`
type filters	no	e.g. `DecisionRecord,Feature`
bundle / repo	no	Default `--repo .`
deepen allowed	no	Default yes; cap is the child's 2 deepen steps

The child walks, scores, and returns only:

## Retrieval card
- Query: …
- Seed: `/path` (`Type`) — why chosen
- Fit: high|medium|low — one sentence
- Engine: index|rg|scan
- Pack: hops=N nodes=N tokens=N/budget
- Lead nodes: (5–8 bullets: title · type · path · one-line why)
- Open gaps: (Questions / missing edges / thin pack) or none
- Next: stay|deepen-2hop|try-alt-seed `/other`

Answer from that card. If the card says deepen-2hop or try-alt-seed and the user still needs more, spawn again with that instruction — still do not pull raw packs into the parent.

Orthogonal fan-out

If the question also needs architecture topology (runtime units, deploy nodes, SAC Package as a build artifact), tell the parent to also spawn SAC architecture-retriever. Do not implement SAC here. PKC Package is project-memory (what shipped).

Done when

The parent has a Retrieval card (or a second card after one re-spawn).
No search hit list or full pack markdown is in the parent context.
The parent has not written knowledge nodes as a side effect of retrieval.

---name: pkc-retrievedescription: Spawn the knowledge-retriever sub-agent to search and pack project memory without contaminating parent context. Use before answering from the OKF tree, or when starting work that needs Feature/Decision context.---# PKC RetrieveParent-facing. The parent must **not** run `pkc_search` / `pkc_pack` itself for retrieval. Spawn the `knowledge-retriever` sub-agent (Claude Code Task/Agent tool, or host equivalent) and consume only the Retrieval card.## Do not pull raw packs into the parent- Do **not** run `pkc_search.py` or `pkc_pack.py` in the parent turn for retrieval.- Do **not** paste search hit lists, mermaid graphs, or ranked pack bodies into the parent working context.- Progressive disclosure (`--tiny`, hop caps) clips the graph; it does **not** isolate the parent. Isolation is the child agent.## SpawnPass the child:| Input | Required | Notes ||-------|----------|-------|| query | one of query / seed | Free-text question or keywords || seed path | one of query / seed | Absolute in-bundle path, e.g. `/features/user-authentication.md` || type filters | no | e.g. `DecisionRecord,Feature` || bundle / repo | no | Default `--repo .` || deepen allowed | no | Default yes; cap is the child's 2 deepen steps |The child walks, scores, and returns **only**:```markdown## Retrieval card- Query: …- Seed: `/path` (`Type`) — why chosen- Fit: high|medium|low — one sentence- Engine: index|rg|scan- Pack: hops=N nodes=N tokens=N/budget- Lead nodes: (5–8 bullets: title · type · path · one-line why)- Open gaps: (Questions / missing edges / thin pack) or none- Next: stay|deepen-2hop|try-alt-seed `/other````Answer from that card. If the card says `deepen-2hop` or `try-alt-seed` and the user still needs more, spawn again with that instruction — still do not pull raw packs into the parent.## Orthogonal fan-outIf the question also needs **architecture topology** (runtime units, deploy nodes, SAC Package as a build artifact), tell the parent to also spawn SAC `architecture-retriever`. Do not implement SAC here. PKC `Package` is project-memory (what shipped).## Done when- The parent has a Retrieval card (or a second card after one re-spawn).- No search hit list or full pack markdown is in the parent context.- The parent has not written knowledge nodes as a side effect of retrieval.

Frequently asked questions