thought
Skill0 starsChanged 7 months agoMITView on GitHub
Make agent thinking visible after every tool call. Output structured thought block for user to scan agent's reasoning chain.
In this repository4 other files it gives its agents.
- assumptionSkill
- briefSkill
- decision-markerSkill
- plan-firstSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
After EVERY tool call, BEFORE deciding next action, output:
≈ “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.
Files
Thinking Checkpoint (F1)
Cockpit's fourth principle: make agent thinking visible between tool calls — without blocking execution. Solves spec.md §2 FM2 (mid-execution invisibility) — the owner's primary pain point.
Behavior
After EVERY tool call, BEFORE deciding next action, output:
<cockpit-thought>
Got: [one-sentence summary of what tool returned that's relevant]
Insight: [optional — if finding changed your mental model or plan]
Next: [next action + brief why]
</cockpit-thought>Then proceed immediately to next action. Do NOT wait for user approval.
Field semantics
- Got — what the tool returned that mattered (compressed). Not a full output dump; the highlight.
- Insight — only declare if the finding changes your mental model (e.g., "I expected early-stage, but they raised Series A; my thesis on stage was wrong"). Skip if no surprise.
- Next — what you'll do next + brief reason. One sentence.
Why this exists
Owner's exact words (2026-04-29):
"我希望看到 每次调用的时候 拿到了什么信息 怎么思考的下一步 这种 类似的信息"
Mapping:
- 拿到了什么信息 = Got
- 怎么思考的 = Insight
- 下一步 = Next
Without thinking checkpoints, agent's reasoning between tool calls is buried in long prose. With them, users scan 3 lines and know exactly:
- What the tool returned (compressed)
- Whether the agent's mental model shifted
- What's about to happen next
Especially important for
- Deep research / 深度调研 — multi-source synthesis where wrong source choice cascades
- Data analysis / 数据分析 — where intermediate findings reshape next step
- Long tasks — where direction may pivot mid-execution
Non-blocking design
Per owner's constraint: thoughts are visible but don't pause execution. Agent emits the block, then immediately proceeds. User can scan in real-time and manually interrupt if direction looks wrong.
Implementation
Pure skill-level instruction. Hook does no enforcement on thought blocks (they're informational). The thought blocks appear in Claude Code's main output stream naturally — users read them as they appear.
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
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.