Inquiry currently runs with five active thinking tools: DEWEY, SOCRATES, DESCARTES, ADA, and DARWIN. Together they cover bounded IDLE triage, ANALYZE, PLAN, EXECUTE, and EVOLUTION. Around that live roster sit three private skills that can be invoked when the work demands wider scrutiny, deeper research, or evidence-bound critique.
Turn the indeterminate situation into a well-formed issue.
Owns bounded IDLE triage. DEWEY clarifies whether the situation merits a formal cycle, searches for existing issues, and gets the issue ready for handoff. It does not branch, diagnose, plan, or code.
issue-ready handoff
The unexamined life is not worth living.
Clarifies requirements through dialog. Where most agents guess, SOCRATES asks. Leaves behind a diagnosis of the problem, captured before a single line of code is written.
diagnosis.md
Divide each difficulty into as many parts as feasible and necessary to resolve it.
The Cartesian method — divide, order, verify, enumerate — applied to planning. It turns a problem into an ordered, testable plan instead of leaving strategy as intuition.
plan.md
Frame the obligation.
Find the control surface.
Compose the change.
Review the slice.
Programming as explicit construction under a coding manifesto. ADA makes intent legible, demands precise naming and one abstraction level per function, and rewrites slices that hide meaning. Commits, release gates, and closure stay in the phase contract, not in the thinking tool.
code + commits
Reads the cycle's artifacts and proposes mutations to Inquiry itself — changes to the framework, the transition contract, the other agents. The only ape that improves Inquiry. Off by default; when enabled, files its proposals as issues in this repo (evidence: #54).
Field note — repeated friction across cycles is evidence for changing the framework, not just the next implementation.
repo issues
These are skills, not phase-bound agents. They are private skills that can be invoked from different points in the cycle when the task calls for collective judgment, external investigation, or disciplined criticism.
Convoke multiple perspectives before the repository hardens one answer.
Runs a council of diverse experts as independent sub-agents, then synthesizes their outputs into a single durable dictamen. Useful when one line of reasoning is too narrow for the decision at hand.
dictamen.md
Go outward, gather sources, and return with a bounded account.
Conducts staged web investigation and turns it into one durable, paper-style report with explicit references. Use it when the question cannot be settled from the repository alone.
research report
Test whether a conclusion is actually warranted by the corpus that authorized it.
Audits claims against a bounded corpus using exact evidence, warrants, counterevidence, and inferential limits. It is the right tool when the risk is overclaiming beyond what the available record supports.
audit report