godmode-repair
Skill6 starsChanged 3 months agoApache-2.0View on GitHub
Re-pitch an answer that did not land, and record why it did not. Use when the operator says they do not follow, asks for clarity, repeats a question already answered, or asks what is being waited on. Not for first answers; do not use when the operator has not signalled confusion.
In this repository22 other files it gives its agents.
- llms.txtllms.txt
- godmode-budgetSkill
- godmode-changelogSkill
- godmode-ciSkill
- godmode-codegraphSkill
- godmode-code-of-lawSkill
- godmode-continuitySkill
- godmode-evidenceSkill
- godmode-governanceSkill
- godmode-host-syncSkill
- godmode-impact-gateSkill
- godmode-investigationSkill
- godmode-loop-wardenSkill
- godmode-memory-gardenerSkill
- godmode-replicateSkill
- godmode-researchSkill
- godmode-second-lookSkill
- godmode-skill-evalSkill
- godmode-skill-forgeSkill
- godmodeSkill
- godmode-spec-lifecycleSkill
- godmode-triageSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 53
Not an apology, and not a re-litigation of whether the first answer was correct.
- Distinctive: line 57
Do not close a repair by asking whether it landed.
≈ “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
Godmode Repair
Outcome
Say the thing again in a form the operator can act on, and leave a record that the first form failed — so the same shape of answer is not produced again next session.
The signal
Repair is invoked by the operator, not by the agent's own sense that things went well. Treat any of these as the trigger, without arguing the point:
- "I don't follow", "be clear", "what do you mean"
- A question already answered, asked again
- "What do you want from me", "what is pending", "are you waiting"
- A decision requested twice without an answer arriving
The last two are specific and worth naming: they mean options were presented where a recommendation was owed. An operator asking what is needed from them is reporting that the ask was buried, not that they missed it.
Re-pitch
- Lead with the answer. One sentence, first line, before any qualification. If the honest answer is "nothing" or "I don't know yet", that is the sentence.
- Cut the options. Where a choice was offered, decide it and say what was decided and why. Reserve the question for the case where the wrong guess is unsafe or wastes the work.
- Name the single next act. Whose it is, and what happens if it does not happen.
- Drop the qualifications the first version carried. They are what buried it.
Length is the usual cause. An answer that needs a table of contents has already failed.
Record it
godmode remember --kind lesson --subject "<what did not land>" --value "<the rule>" --guard "<the shape to
avoid>"
Recorded as a lesson rather than a note because the failure is generalisable: what did not land is almost never this one sentence, it is a shape — options where a decision was owed, a status buried under evidence, a question asked four ways.
Record the shape, not the exchange. The transcript is the host's; a second copy is a second thing to leak.
What repair is not
Not an apology, and not a re-litigation of whether the first answer was correct. It usually was correct and unusable, and those are different faults with different fixes. Restating the reasoning in more detail is the failure repeating itself.
Do not close a repair by asking whether it landed. Say it, then continue the work.
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.