| name | skeptic |
| description | Adversarial critique on demand. Returns the strongest case AGAINST a draft, plan, or decision before it ships. Forbids generic mush. |
| user-invocable | true |
| allowed-tools | Read, Bash, Grep |
Skeptic
Get the strongest case against an artifact before you ship it. The skeptic agent argues the opposite position — counter-thesis, concrete failure modes, hidden assumptions — drawn from the artifact itself, vault counterweights, and past decision-records of similar shape.
When to use
- Before /vault-publish on a non-trivial vault note (Gate 6 fires automatically on hub-importance notes; for lower-stakes notes that still matter, fire manually)
- Before committing a major decision-record
- Before sending an external memo or talk
- Before locking in a plan that has compounding consequences
- Whenever the existing pipeline (researcher → planner → implementer → reviewer) is starting to feel like an echo chamber
What this does
Dispatches the skeptic agent (via the Task tool with subagent_type: skeptic) with $ARGUMENTS as the artifact.
$ARGUMENTS formats:
- File path:
vault/career/development/<slug>.md
- Quoted text:
"the proposition to argue against"
- PR reference:
--pr 88
Output is structured: Counter-thesis / Failure modes / Hidden assumptions. Each item is artifact-specific and concrete; the agent prompt explicitly bans "on the other hand," "it depends," and unspecified-mechanism risks.
You decide if the objections land. The agent argues; you arbitrate.
Arguments: $ARGUMENTS
