agentleFS
AbilityaiAbilityai/cornelius/.claude/skills/think-about-it/SKILL.md

think-about-it

Skill109 starsChanged 39 days agoMITView on GitHub

Free-form thinking mode - consider a topic through successive distinct knowledge-base perspectives (hub lenses, original frameworks, wildcard distant notes). One perspective pass per invocation, state persists across runs, emits [[DONE]] when the exploration weaves together. Loop-compatible by design.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 5 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    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:

    No arguments given → list existing freeform topics with their status and pass counts, exit cleanly (no [[DONE]]).

    line 51
  2. Very specific:

    Read lenses_used from frontmatter.

    line 56
  3. Distinctive:

    Never reuse a lens within a topic.

    line 71
  4. Distinctive:

    Prefer the unused lens with the most *generative friction* - not the most obviously related one.

    line 72
  5. Very specific:

    Read the 3-4 most promising notes in full.

    line 84
  6. Distinctive:

    Write 2-4 paragraphs of actual free-form thinking: What does the topic look like from inside this lens?

    line 88
  7. Distinctive:

    No hypotheses, no confidence percentages - that is the incubation loop's job.

    line 90
  8. Distinctive:

    If a thread contradicts an earlier pass, name the contradiction explicitly and leave it standing.

    line 91
  9. Very specific:

    Update frontmatter: pass_count, lenses_used, last_two_novelty, updated, updated_by.

    line 114

≈ “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

Think About It

Free-form counterpart to the incubation loop. No hypotheses, no confidence tracking, no fixed move rotation - just the same topic viewed through a different KB lens each pass, until the perspectives weave into something worth keeping.

Purpose

Put Cornelius in open thinking mode on a topic. Where /incubation-loop converges on an answer to a question, /think-about-it circles a topic - each pass looks through one distinct perspective drawn from the knowledge base and writes down what that lens reveals, distorts, or contradicts. The exploration ends with a Weave: a synthesis of the threads and candidate permanent-note seeds.

Loop Compatibility Contract

This skill is designed to be driven by a loop (/loop, /trinity:loop, or manual re-invocation):

  • One perspective pass per invocation (single-task rule; each pass completes in well under 10 minutes)
  • All state lives in the topic file - safe across separate context windows; no in-session memory assumed
  • Fixed reply format (Step 7) so {{previous_response}} chaining and Until-mode sentinels work
  • [[DONE]] appears in the reply if and only if the topic file's status is woven - tied to verifiable file state, not self-assessment
  • Idempotent on woven topics - re-invoking a finished topic re-reports the Weave and [[DONE]] without writing anything

State Dependencies

SourceLocationReadWriteDescription
Topic fileBrain/05-Meta/Thinking/Freeform/[slug].md✓✓Perspective passes + frontmatter state
Local Brain Searchresources/local-brain-search/✓Semantic search (--no-track)
Permanent notesBrain/02-Permanent/✓Evidence + wildcard lens source
MOCsBrain/03-MOCs/✓Hub lens grounding

Deliberately NOT in THINKING-REGISTRY.md - freeform topics are invisible to the incubation loop. The two engines never touch each other's state. If a Weave concludes the topic deserves rigorous treatment, promote it manually via /manage-thinking-topics.

Process

Step 1: Resolve Topic and State

$ARGUMENTS = the topic, free text. Slugify to kebab-case.

date '+%Y-%m-%d'
ls "Brain/05-Meta/Thinking/Freeform/" 2>/dev/null
  • File Brain/05-Meta/Thinking/Freeform/[slug].md missing → create it from the template below (pass_count: 0, status: exploring), then continue.
  • No arguments given → list existing freeform topics with their status and pass counts, exit cleanly (no [[DONE]]).
  • File exists with status: woven → reply with the Weave summary and end with [[DONE]]. Write nothing.

Step 2: Pick the Next Lens

Read lenses_used from frontmatter. Choose exactly ONE unused lens from the deck:

Hub lenses (the KB's thematic centers of gravity): Dopamine & Motivation · Flow & Peak Performance · Buddhism & Consciousness · AI Agents & Architecture · Decision-Making & Biases · Identity & Belief Systems · Attention Economics

Framework lenses (the user's original frameworks - view the topic from inside one): Uncertainty-Dopamine-Belief Loop · The Folder Paradigm · Cross-Domain Consilience · External Anchor Principle · Self-Construction Stack · Three Gaps Framework · Four Pillars of Deep Agency · Safety as Velocity

Wildcard lens (auto-discovery style - a random permanent note from an unrelated cluster):

python3 -c "import random,glob; print(random.choice(glob.glob('Brain/02-Permanent/*.md')))"

Re-roll once if the sampled note is obviously about the topic itself; otherwise the friction is the point.

Selection rules:

  • Never reuse a lens within a topic.
  • Prefer the unused lens with the most generative friction - not the most obviously related one. Obvious relevance produces summary; friction produces thought.
  • By pass 4 at least one wildcard must have been used; if not, this pass is a wildcard.

Step 3: Gather KB Material Through That Lens

# --no-track: autonomous reads must not train q-values (learning hygiene, same as incubation-loop)
# read role: reasoning (contract: scope-mount) - the reasoning mount, so a lens can land on what the user has READ, not only what he wrote
BRAIN_READ_SCOPE=core,Books,document-insights python3 resources/local-brain-search/search.py "[topic] [lens core terms]" --limit 6 --mode spreading --no-track 2>/dev/null
BRAIN_READ_SCOPE=core,Books,document-insights python3 resources/local-brain-search/search.py "[lens core concept]" --limit 4 --mode spreading --no-track 2>/dev/null

Read the 3-4 most promising notes in full. These ground the pass - every pass must cite real notes, and a Books/ or Document Insights/ note is cited as encountered material, never as the user's view.

Step 4: Think

Write 2-4 paragraphs of actual free-form thinking: What does the topic look like from inside this lens? What does this lens make visible that previous passes missed? What does it distort or refuse to see? Capture analogies, tensions, reframings, open questions.

  • No hypotheses, no confidence percentages - that is the incubation loop's job.
  • If a thread contradicts an earlier pass, name the contradiction explicitly and leave it standing. Premature resolution kills the weave.
  • Preserve the user's framing vocabulary where the notes supply it.

Step 5: Append the Pass

Append to the topic file:

### Pass [N]: Through the lens of [Lens Name] - YYYY-MM-DD

**Notes consulted:** [[Note A]], [[Note B]], [[Note C]]

**What this lens reveals:**
[2-4 paragraphs from Step 4]

**Threads pulled:**
- [tension / analogy / question worth keeping - 1 line each]

**Novelty:** high | medium | low

Rate novelty honestly: high = a genuinely new thread appeared; medium = deepened an existing thread; low = mostly re-derived prior passes. The honest low is what lets the loop terminate.

Update frontmatter: pass_count, lenses_used, last_two_novelty, updated, updated_by.

Step 6: Weave Check

Weave when EITHER:

  • pass_count >= 7, OR
  • the last two passes are both novelty: low (the lenses have stopped paying)

If weaving, append:

---

## The Weave

[3-6 paragraphs braiding the threads: the strongest cross-lens pattern, the contradiction
that stayed productive, what the topic turned out to actually be about]

**Candidate permanent-note seeds:**
- [2-3 one-line seeds that could become atomic notes]

**Promote to incubation?** [yes + suggested central question | no + why not]

Set status: woven in frontmatter.

Step 7: Reply (Fixed Contract)

  • Not woven: Pass N - lens: [lens]. [The single most interesting thread, 1-2 sentences.] Continuing.
  • Woven: 2-3 sentence Weave summary + the candidate note seeds + final line [[DONE]]

The reply ends with [[DONE]] if and only if status is woven.

Looping Recipes

Remote (Trinity), Until mode - the demo shape:

/trinity:loop git pull --rebase, then run /think-about-it "<topic>", then commit and push;
relay the skill's reply verbatim — until done, max 10, every 30s

(stop_signal: [[DONE]], max_runs: 10, delay_seconds: 30, timeout_per_run: 1200)

Local dynamic loop: /loop run /think-about-it "<topic>"; stop when the reply contains [[DONE]]

Manual: just invoke /think-about-it "<topic>" repeatedly; each call advances one pass.

Topic File Template

---
created: YYYY-MM-DD
updated: YYYY-MM-DD
created_by: [model-name]
updated_by: [model-name]
agent_version: 01.25
topic: "[topic as given]"
pass_count: 0
lenses_used: []
last_two_novelty: []
status: exploring
---

# Thinking Freely: [Topic Title]

## Topic
[The topic as given, one sentence of framing if needed]

---
*[Perspective passes appended below by /think-about-it]*

Error Handling

ErrorRecovery
Freeform/ directory missingCreate it, continue
KB search returns emptyRetry with alternate terms from the lens; if still empty, ground the pass in the wildcard note alone and say so in the pass
All lenses in the deck used before weaveForce weave this pass regardless of novelty
Topic file corrupt / unparseable frontmatterRename to [slug]-corrupt-YYYY-MM-DD.md, recreate fresh, note the reset in the new file

Completion Checklist

  • Topic file resolved (created or loaded)
  • One unused lens selected per the rules
  • KB searched through the lens; 3-4 notes read in full
  • Pass appended with notes cited and honest novelty rating
  • Frontmatter state updated
  • Weave check performed; Weave written if triggered
  • Reply follows the fixed contract ([[DONE]] iff woven)

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.