Decide
Open-source AI Operating System (AIOS) blueprint for Claude Code & Codex — turn any agent into your personal second brain + Chief of Staff: skills, self-verifying workflows, a second-brain wiki, and deterministic tools. Fork it for any role or domain.
npx -y skills add alirezarezvani/gaios --skill decideAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
What its author says it does
Copied from the file, not written here
Use when the user faces a real decision with stakes and wants it framed cleanly — options, ranked criteria, a recommendation, the falsifier, and reversibility — then logged to decisions/log.md. Trigger on "/decide", "help me decide", "should I X or Y", "frame this decision", "what's the call here", "log this decision".
SKILL.md
2.4 KB, as published. Nobody here has run it
Decide
Turn a fork in the road into a clear, logged decision — reasoning visible, and the way you'd know you were wrong stated up front.
When to run
- A real choice with stakes (a tradeoff, a hire, a vendor, a direction, a bet).
- After
/structuresurfaces a "Decision needed."
The output (always this shape)
## Decision — <the question> · <date>
**Context** — why this is on the table now (2–3 lines).
**Options** — each with its main upside + main cost/risk (include "do nothing").
**Criteria** — what actually matters here, ranked.
**Recommendation** — the call + the why, in one paragraph.
**Falsifier** — what we'd later see that would prove this was the wrong call (Verification Gate).
**Reversibility** — one-way or two-way door? (two-way → decide fast and move).
**Owner / next step** — who does what, by when.
Process
- Frame the real question. Strip to the actual choice (
/structureif fuzzy). Name what is not being decided. - Lay out options honestly — including "do nothing." Each gets a real upside and a real cost. No strawmen.
- Make criteria explicit and ranked — most decisions go wrong because the criteria were never named.
- Recommend. Take a position with the reasoning. State the falsifier and the reversibility (two-way doors don't deserve one-way-door deliberation).
- Verify the inputs (Verification Gate). Any fact or number the decision rests on → trace to source; flag unknowns as risks, don't paper over them.
- Log it. On the user's confirmation, append the brief to
decisions/log.md(dated) — the standing record (per "How you work with me").
Autonomy
L1–L2 — frames and recommends; the human makes the call. Logs it once the user confirms.
Guardrails (from CLAUDE.md)
- Cite, don't invent — decisions rest on verified facts; unknowns are named as risks.
- Log decisions to
decisions/log.md. - No confidential figures or PHI/PII in the log (it's committed) — reference the source, don't transcribe.