agentleFS
flonatflonat/flonat-research/skills/ideas/SKILL.md

ideas

Skill150 starsChanged 10 days agoMITView on GitHub

Capture and integrate improvement ideas for the shared Claude Code and Codex infrastructure. Use when recording an infrastructure enhancement for later review rather than implementing it immediately.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read log/.ideas-state.json to find the last integration date.

    line 85
  2. Very specific:

    Read log/ideas.md and collect all entries with Status: unprocessed (optionally filtered by date).

    line 89
  3. Distinctive:

    Ask: "Which ideas should I act on?

    line 113
  4. Distinctive:

    Investigations: Create a vault task in the Tasks Tracker with:

    line 118
  5. Very specific:

    Update each processed entry's status in log/ideas.md:

    line 126
  6. Very specific:

    When dryrun is specified, run steps 1-4 only — classify and present, but don't execute or update state.

    line 144

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Ideas: Capture & Integrate

An inbox for improvement ideas — the gap between [LEARN] tags (immediate corrections) and skill-extract (full skill creation). Captures raw improvement ideas and periodically integrates the best ones into the system.

Two Modes

ModeTriggerWhat it does
Captureideas capture [text]Append a structured idea entry to log/ideas.md
Integrateideas integrateReview unprocessed ideas and turn them into actions

Mode 1: Capture

With argument: ideas capture [text]

Parse the text and append a structured entry to log/ideas.md.

Without argument: ideas capture

Ask interactively:

  1. "What's the idea?" (free text)
  2. "Where did this come from?" (current session / external repo / user feedback / reading)

Entry Format

## [YYYY-MM-DD] [category] Idea title
- **Quality:** [high] | [medium]
- **Source:** [session | external:repo-name | user-feedback | reading:source]
- **Proposed action:** One-sentence description of what to do
- **Details:** [the idea text, expanded if needed]
- **Status:** unprocessed

Auto-Classification

Classify each idea into one of these categories based on content:

CategorySignal words / patterns
[skill-design]"new skill", "workflow for", "automate", "slash command"
[rule]"always", "never", "enforce", "prevent", "require"
[workflow]"process", "steps for", "how to handle", "protocol"
[infrastructure]"hook", "MCP", "settings", "permissions", "symlink"
[research]"method", "analysis", "paper", "experiment", "data"

Quality Rating

RatingCriteria
[high]Specific, actionable, addresses a known friction point or recurring issue
[medium]Interesting but needs more thought, or addresses a minor convenience issue

Skip [low] ideas — if it's not worth recording, don't record it.

After Capture

Confirm: "Idea captured in log/ideas.md. You have [N] unprocessed ideas — run ideas integrate when ready to act on them."


Mode 2: Integrate

Trigger: ideas integrate [since:YYYY-MM-DD] [dryrun]

Step 1: Read State

Read log/.ideas-state.json to find the last integration date. If since: argument is provided, use that instead. If no state file exists, process all entries.

Step 2: Gather Unprocessed Ideas

Read log/ideas.md and collect all entries with **Status:** unprocessed (optionally filtered by date).

Step 3: Classify Each Idea

For each unprocessed idea, classify as:

TypeDescriptionAction
Direct proposalSpecific edit to a known file — can be implemented nowPresent the edit for approval
InvestigationNeeds research or design before implementationCreate a vault task
DuplicateAlready covered by an existing skill, rule, or conventionMark as duplicate with reference
DeclinedNot worth pursuing after reflectionMark as declined with reason

Step 4: Present for Approval

Show a summary table:

| # | Idea | Category | Quality | Proposed action | Classification |
|---|------|----------|---------|-----------------|---------------|
| 1 | ... | [rule] | [high] | ... | Direct proposal |
| 2 | ... | [skill-design] | [medium] | ... | Investigation |

Ask: "Which ideas should I act on? (all / numbers / skip)"

Step 5: Execute Approved Ideas

  • Direct proposals: Make the edit, following break-the-glass rules for protected files
  • Investigations: Create a vault task in the Tasks Tracker with:
    • Task name: "Investigate: [idea title]"
    • Source: "Ideas inbox"
    • Priority: High for [high] quality, Medium for [medium]
    • Description: Full idea text + proposed approach

Step 6: Update State

Update each processed entry's status in log/ideas.md:

  • **Status:** implemented (YYYY-MM-DD) — direct proposal was executed
  • **Status:** task-created (YYYY-MM-DD) — vault task created for investigation
  • **Status:** duplicate → [reference] — already exists
  • **Status:** declined — [reason] — not worth pursuing

Write log/.ideas-state.json:

{
  "last_integration": "YYYY-MM-DD",
  "total_captured": N,
  "total_implemented": N,
  "total_pending": N
}

Dry Run Mode

When dryrun is specified, run steps 1-4 only — classify and present, but don't execute or update state. Useful for reviewing the backlog without committing to action.


Important Notes

  • Ideas are not tasks — they're raw material that may or may not become actionable
  • The capture threshold is deliberately low — if something seems worth noting, capture it
  • Integration should happen periodically (weekly or when the backlog reaches 10+ ideas)
  • This skill complements [LEARN] tags (which record corrections) and skill-extract (which creates full skills)
  • Break-the-glass rules still apply — direct proposals that modify protected files need confirmation

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.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

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.