decide-card
Skill7 starsChanged 6 months agoMITView on GitHub
Record a decision (what + why) on a parked card and lower the human gate from decision or session to none. AUTO-INVOKE on "I decided X", "let's go with Y", "approved", or any resolution of a parked card — the human's one-action handoff so pull-card can resume.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- site/llms.txtllms.txt
- advance-cardSkill
- audit-deckSkill
- card-schemaSkill
- claude-kickoffSkill
- create-cardSkill
- deckSkill
- finish-cardSkill
- kickoffSkill
- next-cardSkill
- pull-cardSkill
- refine-deckSkill
- retrospectiveSkill
- scan-deckSkill
- standupSkill
- tune-cadenceSkill
- upgradeSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
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
- Very specific: line 98
If the decision reverses or re-scopes a verdict the card already states, goc decide only writes the ## Decision block — every other surface …
- Distinctive: line 118
Cite the primary source: when the answer rests on literature: DOI/PMID, textbook chapter, or project-doc section.
- Distinctive: line 123
Default to human when ambiguous: If the rubric doesn't give a clean answer, raise the gate.
- Distinctive: line 131
Card needs implementation, not a decision: If you're ready to do the work yourself, just do the work.
- Very specific: line 143
Does NOT implement the card (that's pull-card's next round).
- Very specific: line 144
Does NOT close the card (that's Skill(finish-card)'s DoD-gated contract).
- Very specific: line 146
Does NOT mutate status, except to release a parked claim (active → open, above).
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
When to invoke
Invoke when the user says "I decided X", "let's go with Y", "the answer is Z", "go ahead with", "approved", or otherwise resolves a parked card. The Andon-cord lowering action — pull-card raises the gate; this skill is the human's one-action handoff so pull-card can resume.
Decide a card
The Andon-cord loop: Skill(pull-card) raises human_gate
(none → decision / session) when it hits a question only a human
can answer; this skill lowers the gate back to none in one cheap
action. The card comes out open — one parked while active
(pull-card claims before it works) is released back to the queue —
so the next pull-card claims and implements per the recorded
decision. Humans resolve the cause;
agents wield the wrench — and an agent that can cite the project
rubric MAY decide on its own behalf (lazy Andon, below).
Edge cases live in reference.md — a sibling file in this
skill's directory. Read the named section only when the situation
actually applies:
| Situation | reference.md section |
| Presenting labeled options via AskUserQuestion | Mirroring labeled options |
| The decision reverses / re-scopes a stated verdict | Re-scope reconciliation |
| An agent recorded a decision it shouldn't have | Rewinding an agent decision |
| Why the cord must be cheap to lower | Rationale |
What this skill does to the card
Body. Replaces the
## Decision requiredsection (or appends a fresh## Decisionsection) with:## Decision *Resolved YYYY-MM-DD:* <one-line decision> *Reasoning:* <one-line why>- log.md. Appends up to two entries, timeline-ordered: first an
archive of the prior
## Decision requiredcontent (the options and trade-offs the README rewrite would otherwise lose), then the resolution (<decision> — <reason>. Gate <prior> → none.). - Frontmatter. Flips
human_gate: decision(orsession) →none. Anopencard staysopen. Anactivecard (parked mid-work) returns toopenwith itsworkerkept, becauseactive+ gatenonereads as a live claim that no queue lists; to keep working it in the same session, re-claim it withgoc status <title> active. - Refuses (exit 2, nothing written) if the gate is already
none, the title doesn't exist, or--decision/--becauseis blank.
Workflow
Read the card body (
goc show <title>). Confirm the decision actually answers the parked question: if## Decision requiredenumerates options A/B/C, the user's free-text decision should map onto one (or explicitly reject all and propose D). When eliciting the pick with AskUserQuestion, mirror the body's option labels and order verbatim — do not reorder or rephrase (reference.md§ Mirroring labeled options).Then check
advanced_byfor unclosed prerequisites, and read any you find before recording. Anadvancesedge does not record whether it is strict, so the only way to learn that a prerequisite reframes this card is to open it. Lowering the gate is what makes the card autonomously pullable, so a decision taken over an unread prerequisite is one an unattended worker may act on before any human sees the card again.goc decideprints the same advisory (non-blocking — it names the prerequisites and lowers the gate anyway), but it prints it as the decision lands, which is too late to inform it.Run the CLI.
goc decide <title> \ --decision "<one-line what>" \ --because "<one-line why>"- Commit follows repo policy.
goc decidehonors.game-of-cards/config.yamlworkflow.auto_commit(defaulttrue), committing body + log + gate flip asdecide: <title> — <decision>. Override per invocation with--no-commit/--commit.
If the decision reverses or re-scopes a verdict the card already
states, goc decide only writes the ## Decision block — every
other surface still asserting the old verdict is now stale. Reconcile
by hand: the summary: frontmatter, any > ⚠ body banner, DoD
wording, and mentions in neighbor cards. The CLI prints a reminder on
re-scope language and goc validate flags
DECISION_CONTRADICTS_VERDICT. For a true re-scope, prefer
supersede + create over an in-place rewrite. Full checklist:
reference.md § Re-scope reconciliation.
When an agent invokes this skill (lazy Andon)
An agent may decide on its own only after consulting the consuming repo's project-specific rubric:
!cat .game-of-cards/hooks/decide-card.md 2>/dev/null || true
- Cite the rubric.
--becauseMUST start with a clause naming the rubric consulted and what it returned (citation form per the hook above). - Cite the primary source when the answer rests on literature: DOI/PMID, textbook chapter, or project-doc section.
- No agent-decide for human-judgement questions — resource allocation, scope splits, deadlines, stakeholder alignment, taste calls. Raise the gate instead.
- Default to human when ambiguous. If the rubric doesn't give a clean answer, raise the gate; don't guess.
A decision recorded in violation of this contract can be rewound —
reference.md § Rewinding an agent decision.
When NOT to use this skill
- Card needs implementation, not a decision. If you're ready to do the work yourself, just do the work.
- Decision changes scope. "Scope this differently / split /
reframe" is not a resolution — use
Skill(advance-card) <title> superseded+Skill(create-card)for the replacement. - Gate is
sessionand the session never happened. Capturing a real session's outcome is valid; shortcutting a session that never took place is not — schedule the session instead.
What this skill does NOT do
- Does NOT implement the card (that's
pull-card's next round). - Does NOT close the card (that's
Skill(finish-card)'s DoD-gated contract). - Does NOT mutate
status, except to release a parked claim (active→open, above).
Cross-references
reference.md(this skill's directory) — edge cases routed above.Skill(scan-deck)— surfaces parked cards; its "decisions to make" Q&A calls this skill per card.Skill(advance-card)— status machine +waiting_onoverlay; gate ≠ status.Skill(card-schema)—human_gateenum and the## Decision requiredbody convention.
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.

