agentleFS
hsigstadhsigstad/research-kit/skills/idea/SKILL.md

idea

Skill0 starsChanged 4 months agoMITView on GitHub

Create a new research idea with YAML frontmatter and add it to ideas/index.md. Use when the user wants to record a new research idea.

27 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use weekly-retro, decompose and 7 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Joshua-PalamuttamSkill

    weekly-retro

    End-of-week retrospective. Compares planned vs actual, updates calibration data, and surfaces estimation patterns. Run Friday afternoons.

    Also in 3 projects that use this

  • matt-k-wongSkill

    decompose

    Learn to decompose complex problems before building real agents. Better answers today. Ready for orchestration tomorrow. Breaks messy tasks into clear DAG structures with explicit dependencies, parallel branches, and synthesis logic. Improves Claude outputs by enforcing topological order and catching contradictions. Teaches the mental model for LangGraph, CrewAI, n8n, and Sub-Agents. Type "decompose: [goal]" or "decompose preview: [goal]" to start. Zero experience required.

    Also in 3 projects that use this

  • seongjaeryuSkill

    save

    Complete and save your current thinking to Bindly

    Also in 3 projects that use this

  • AWSSkill

    retro

    Conduct a retrospective on a coding session.

    Also in 5 projects that use this

  • GoogleSkill

    to-issues

    Break a plan, spec, or PRD into independently-grabbable issues on the project issue tracker using tracer-bullet vertical slices.

    Also in 3 projects that use this

  • GoogleSkill

    to-prd

    Turn the current conversation into a PRD and publish it to the project issue tracker — no interview, just synthesis of what you've already discussed.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

10 rules. 0 are common, 5 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 22 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Create a research idea file with proper frontmatter and register it in the index.

    line 8
  2. Distinctive:

    Ask the user (if not already provided):

    line 16
  3. Very specific:

    Add a row to the appropriate table in $ROOT/ideas/index.md:

    line 48
  4. Distinctive:

    After creating the idea, briefly check:

    line 62
  5. Very specific:

    When an idea graduates to projects/<slug>/, the exploratory source//build/ is the seed of the real project tree — move and re-home it under …

    line 109
  6. Very specific:

    Update the index link from <slug>.md to <slug>/summary.md

    line 123
  7. Very specific:

    Update any internal links inside summary.md that referenced the old paths

    line 124
  8. Very specific:

    Use existing tags where possible.

    line 132
  9. Distinctive:

    If the idea is clearly connected to an existing idea, mention it but don't merge them.

    line 136
  10. Distinctive:

    If the user is recording an idea that already has substantial content (literature notes, evaluation, etc.), consider creating it directly …

    line 137

“Common” is measured: kept by at least 6 in 10 of the 22 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

New Research Idea

Create a research idea file with proper frontmatter and register it in the index.

Finding the workspace root

The workspace root contains CLAUDE.md alongside projects/, pipelines/, ideas/, research/. If the current directory is inside a project or pipeline, search upward to find the root. Use $ROOT for all paths below.

Step 1: Gather information

Ask the user (if not already provided):

  1. Title — descriptive name for the idea
  2. Rank — A (high priority), B (medium), C (low), or null (unranked)
  3. Tags — 2-3 topic tags (check existing tags in $ROOT/ideas/index.md for consistency)
  4. Content — the idea itself (can be brief)

Optionally ask:

  • Related project — if connected to an existing project under projects/
  • Related ideas — if it connects to existing ideas (check $ROOT/ideas/index.md)

Step 2: Create the idea file

Derive a slug from the title (lowercase, hyphens, e.g., optimal-appeals-system.md).

Write $ROOT/ideas/<slug>.md:

---
title: <Title>
rank: <A/B/C or null>
status: idea
tags: [<tag1>, <tag2>]
related_project: <projects/xxx or null>
last_updated: <today's date>
---

<Content — the idea description, motivation, possible approaches, etc.>

Step 3: Update the index

Add a row to the appropriate table in $ROOT/ideas/index.md:

  • Ranked ideas table (if rank is A, B, or C) — insert in rank order, then alphabetically within rank:
    | <rank> | [<title>](<slug>.md) | idea | <tag1>, <tag2> |
  • Unranked ideas table (if rank is null):
    | [<title>](<slug>.md) | idea | <tag1>, <tag2> |

Step 4: Suggest connections

After creating the idea, briefly check:

  • $ROOT/ideas/index.md for related existing ideas (by tag overlap or topic)
  • $ROOT/research/meta/research_map.md for connections to existing projects
  • Mention any connections to the user but don't modify other files.

Promotion to a folder (proto-project structure)

New ideas always start as flat <slug>.md files. When an idea accumulates more than its description — literature notes, evaluation documents, design notes, lit search artifacts — promote it to a folder. The promoted layout is:

ideas/<slug>/
  summary.md       ← was <slug>.md, with frontmatter
  synthesis.md     ← literature read, if any
  evaluation.md    ← assessment as policy/theory
  todo.md          ← next steps, decision rule, time budget
  design.md        ← detailed design notes
  paper.tex        ← paper sketch once idea has crystallized
  references.bib   ← bibliography for paper.tex
  literature/      ← lit search artifacts (queries.json, candidates.json, etc.)
  source/          ← exploratory analysis scripts (optional; see below)
  build/           ← outputs of those scripts (optional; see below)

Exploratory source/ and build/ in an idea folder

An idea folder may carry source/ and build/ subdirectories to hold exploratory analysis run before the idea is promoted to a proper project — the point is to probe feasibility (does the data support the design? is there a sample? is the effect visible?) without paying the cost of a full projects/<slug>/ scaffold.

This is a sanctioned sandbox, deliberately lighter than the project docs contract:

  • The doc-contract linter (check_docs.py) does not scan inside idea folders — find_repos only walks projects/* and pipelines/*, and lint_ideas only checks flat ideas/*.md frontmatter. So the source/X.py → build/X.ext bijection, required docs, and merge-validate= rules are not enforced here. Keep the folder tidy by convention, not by hook.
  • Still follow the light habits that make the work legible: mirror source/X.py → build/X.<ext>, keep a short INTENT line at the top of each script (Inline Audit Trail), and read only cleaned pipeline outputs (pipelines/*/build/), never raw data.
  • When an idea graduates to projects/<slug>/, the exploratory source//build/ is the seed of the real project tree — move and re-home it under the full docs contract at that point.

Record what the exploration found in summary.md (or a findings.md in the folder) so the conclusion survives even if the scratch scripts are later pruned.

To promote a flat idea to a folder:

  1. Create ideas/<slug>/
  2. Move <slug>.md → <slug>/summary.md
  3. Move any related material (e.g., from ideas/literature/<slug>/) into the new folder
  4. Update the index link from <slug>.md to <slug>/summary.md
  5. Update any internal links inside summary.md that referenced the old paths

The /idea skill itself only creates flat files. Promotion is a manual step taken when an idea has clearly outgrown a single document.

Gotchas

  • Use existing tags where possible. Check $ROOT/ideas/index.md for the tag vocabulary already in use before inventing new tags.
  • The slug should match the title closely — look at existing files for the naming convention (e.g., effects-of-courts-on-corruption.md).
  • Status is always idea for new ideas. Other values (exploring, shelved, project) are set later.
  • For small notes that don't warrant their own file, suggest adding to an existing topic collection in $ROOT/ideas/topics/ instead.
  • If the idea is clearly connected to an existing idea, mention it but don't merge them.
  • If the user is recording an idea that already has substantial content (literature notes, evaluation, etc.), consider creating it directly as a folder instead of a flat file.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.