agentsclimarketplace

Decision review

Skill arniesaha/ctxops/skills/decision-review

Agent skills (open standard) for the bookends of the agent loop: context assembly, decision review, and work-shaping outputs. Works with Claude Code and any Agent Skills harness. Ships its own adoption and value measurement.

Install
npx -y skills add arniesaha/ctxops --skill decision-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

WHEN about to commit to a direction and want a sharp adversarial pass. Triggers: 'should we do X?', 'is this worth it?', 'pressure-test this plan/principle/vision', 'pick between A and B', 're-check the call we made two weeks ago', 'play devil's advocate'. Direct, not balanced. OUTPUT: verdict (do/delay/narrow/compare/kill), strongest case for and against, cheapest test. PAIR BEFORE: ticket-scoping (file the chosen path). Use post-hoc mode for older calls.

SKILL.md

1.8 KB, as published. Nobody here has run it

decision-review

Adversarial review of a single direction or choice between a few. This culls existing options; it does not generate new ones.

Workflow

  1. Restate the decision in one sentence.
  2. Identify the actual constraint: time, money, trust, focus, technical risk, market risk, or reversibility.
  3. Make the strongest case for the direction.
  4. Make the strongest case against it.
  5. List hidden assumptions.
  6. Identify the cheapest test.
  7. Recommend: do, delay, narrow, compare, or kill.

Post-hoc Mode

When the user asks whether a past call still holds (e.g. "we decided X two weeks ago - is that still right?"):

  1. Recover the original decision and its assumptions from memory / repo / docs.
  2. List which assumptions are now confirmed, broken, or unknown.
  3. Re-run steps 3-7 with current facts.
  4. Recommend: keep, revise, or reverse - and the smallest action to lock it in.

Output Shape

  • Decision
  • Verdict
  • Strongest case for
  • Strongest case against
  • Hidden assumptions
  • Cheap test
  • What would change the verdict

Rules

  • Be direct. Do not balance weak arguments for politeness.
  • If the real issue is positioning, say so.
  • If the real issue is execution capacity, say so.
  • Distinguish reversible choices from one-way doors.
  • Recommend the smallest action that creates evidence.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.