How do I install the Groundwork Seed skill?

Run npx skillmds@latest add glebis/groundwork-seed 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 Groundwork Seed 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 Groundwork Seed?

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 Groundwork Seed free to use?

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

Who published Groundwork Seed?

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

Groundwork Seed by glebis — Agent Skill

Name: Groundwork Seed
Author: glebis

File contents

groundwork-seed

Reads distilled research findings and proposes an initial draft of the target artifact. For v1, values is the only target.

When to use

After groundwork-ingest produced a non-empty research-findings.md.
The user wants to warm the artifact before running a real session.

What you do

Read context. Profile, research-findings.md, frameworks with artifacts_served: [values].
Propose candidates. From the declared + implied values in research-findings, plus any tensions, propose 5-7 candidate values. For each:
- A short phrase (3-7 words — "truth over comfort," "kin over scale")
- A one-line rationale grounded in one or more source citations
- A tag: declared (user said this explicitly) or implied (pattern-inferred)
Present to user as a review list. Format:

Values candidates seeded from your material:

  1. truth over comfort           [declared · sources: s03, s07]
     "I'd rather be right and alone than agreeable and wrong."

  2. kin over scale               [implied · sources: s01, s02, s05]
     You return often to friction over scale-first thinking.

  3. …

Mark each: keep (k), cut (c), or sit-with (s).

Ask the user to mark each. Interactive, not headless.

Merge into the artifact. For each kept: add to ## Candidate values with the rationale and source slugs. For each sit-with: add under a ## Sitting with subsection. Cut ones are recorded in the Changelog as "considered and rejected" but not in the main list.
Update frontmatter. Set status: drafting. Regenerate speakable_summary (40 words). Do not set last_synthesized — seeding is not synthesis.
Log the seed. Write sessions/{YYYYMMDD}-{HHMM}-values-seed.md with the user's keep/cut/sit-with decisions (for audit).

Tiers

line: "seeded {n} candidates"
brief: the count + the kept list
full: the kept + sit-with + cut lists with rationales

Failure modes

research-findings.md empty → stop; tell user to run groundwork-ingest first.
User declines all candidates → that's a signal, not an error. Log the seed session anyway; suggest a real session.

---name: groundwork-seeddescription: Read research-findings.md and frameworks, propose an initial draft of an artifact (Values in v1), confirm with user, merge into the artifact file. Eliminates the blank-page problem.---# groundwork-seedReads distilled research findings and proposes an initial draft of the target artifact. For v1, `values` is the only target.## When to use- After `groundwork-ingest` produced a non-empty `research-findings.md`.- The user wants to warm the artifact before running a real session.## What you do1. **Read context.** Profile, research-findings.md, frameworks with `artifacts_served: [values]`.2. **Propose candidates.** From the declared + implied values in research-findings, plus any tensions, propose 5-7 candidate values. For each: - A short phrase (3-7 words — "truth over comfort," "kin over scale") - A one-line rationale grounded in one or more source citations - A tag: `declared` (user said this explicitly) or `implied` (pattern-inferred)3. **Present to user as a review list.** Format:```Values candidates seeded from your material: 1. truth over comfort [declared · sources: s03, s07] "I'd rather be right and alone than agreeable and wrong." 2. kin over scale [implied · sources: s01, s02, s05] You return often to friction over scale-first thinking. 3. …Mark each: keep (k), cut (c), or sit-with (s).```Ask the user to mark each. Interactive, not headless.4. **Merge into the artifact.** For each kept: add to `## Candidate values` with the rationale and source slugs. For each sit-with: add under a `## Sitting with` subsection. Cut ones are recorded in the Changelog as "considered and rejected" but not in the main list.5. **Update frontmatter.** Set `status: drafting`. Regenerate `speakable_summary` (40 words). Do not set `last_synthesized` — seeding is not synthesis.6. **Log the seed.** Write `sessions/{YYYYMMDD}-{HHMM}-values-seed.md` with the user's keep/cut/sit-with decisions (for audit).## Tiers- **line**: `"seeded {n} candidates"`- **brief**: the count + the kept list- **full**: the kept + sit-with + cut lists with rationales## Failure modes- `research-findings.md` empty → stop; tell user to run `groundwork-ingest` first.- User declines all candidates → that's a signal, not an error. Log the seed session anyway; suggest a real session.

Frequently asked questions

Groundwork Seed