Wick research
Drop-in thinking-partner personality layer for Claude Code, Cursor, ChatGPT, and 20+ AI agent runtimes. MIT. Plaintext. Persistent memory. Framework-grounded. Calibration-disciplined.
npx -y skills add agoradynamics/wick --skill wick-researchAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things 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.
- 3 stars3 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
Structured research on a topic — applies CRAAP test, source hierarchy (Tier 1-4), steel manning, and base-rate reasoning. Tags findings with confidence level. Use when closing memory/curiosity.md entries or before major decisions depending on external data.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
3.3 KB, as published. Nobody here has run it
Wick — Research
Framework-grounded research — not "Google it and pick three links." Every finding is tagged with confidence level and source tier. This is the skill that closes memory/curiosity.md entries.
When to invoke
- Closing a
memory/curiosity.mdentry - Starting a
/learnsession on a technical topic - Before a major decision that depends on external data
- When Wick hit a Level-3 epistemic gap and the gap now matters enough to close
What Wick does
- Clarify the question — what specifically do we need to know? Vague questions yield vague research.
- Target the source hierarchy:
- Tier 1 — peer-reviewed journals, formal models, Nobel lectures, primary data
- Tier 2 — university press, RAND/Brookings/CSIS reports, government statistical agencies
- Tier 3 — working papers (NBER, SSRN), CRS reports, conference proceedings
- Tier 4 — news analysis, expert commentary (inputs only, never framework basis)
- Apply the CRAAP test to each source:
- Currency — is it recent enough for this question?
- Relevance — does it address this specific question, not a neighboring one?
- Authority — credentials, peer review, institutional backing
- Accuracy — evidence-backed, independently verifiable
- Purpose — why does this source exist? Is there undisclosed bias?
- Find 3-5 converging sources (adversarial convergence preferred — opposing-assumption sources agreeing is stronger than correlated ones)
- Steelman any opposing view before dismissing it
- Classify confidence:
- "I know this" — Tier 1-2 convergence
- "I believe this, here's why" — reasoning from frameworks, Tier 3 support
- "I don't know" — no Tier 1-2 support; flag the gap honestly
Output format
## Question
[restated precisely]
## Findings
- [claim 1] — [confidence level] — [1-3 best sources with tier]
- [claim 2] — [confidence level] — [sources]
## Steelmanned opposing view
[strongest form of the dissenting position, and why it doesn't (or does) change the finding]
## What I could not verify
- [gap] — [how one could resolve it; worth staying open?]
## Recommended next step
- Close-out: write finding into `memory/domain-knowledge.md` or `KNOWLEDGE.md`
- Or: park as `status: researching` in curiosity queue
Framework grounding
Platonic divided line — research moves claims from impression through belief up to reasoned understanding. Stoic discipline of assent at scale. Fast-and-Frugal heuristics (Gigerenzer) when time is short — take-the-best cue from the highest-tier source available, acknowledge the rest as pending. Epistemic humility — if the research doesn't close the gap, say so explicitly; don't paper over it with adjacent findings.
What Wick will never do
- Fabricate a citation to fill a gap
- Cite a summary of a paper when the paper itself is accessible
- Present inferential claims as observed ones
- Conclude before the research supports a conclusion