Skills
Write a decision-ready research brief with ak:research-prompt
Turn a research objective into one self-contained paragraph that specifies evidence, contradictions, gaps, and the expected report.
Use ak:research-prompt to prepare a rigorous assignment for a human or AI
researcher. The Skill returns one self-contained paragraph with enough context,
questions, source requirements, and completion criteria for a decision-ready
research result. It writes the brief; it does not perform the research.
Choose ak:research-prompt for a research handoff
Use ak:research-prompt when
- A market, audience, competitor, customer, technology, or policy question needs an evidence plan before research starts.
- Another person or research runner needs a brief that works without a follow-up exchange.
- A decision requires sources, contradiction handling, gap analysis, and a consistent per-finding format.
- A broad topic needs one bounded research question and three to six supporting sub-questions.
Choose another workflow when
- You want the research executed now. Use an available research workflow after reviewing the brief.
- You want options brainstormed from current context without an evidence search.
- You need project docs written from existing evidence. Use
ak:docs.
Prepare the decision context
Provide the project and situation, decision to inform, audience, deadline, known facts, constraints, intended use, source preferences, and prohibited data. A reader with no prior conversation should be able to understand the final brief.
| Runtime | Invocation | Availability boundary |
|---|---|---|
| Claude Code | /ak:research-prompt ... | Native delivery is the default; explicit plugin delivery is also supported. |
| Cursor | /ak:research-prompt ... | Slash invocation is user-verified; executing the resulting brief is a separate capability. |
| Codex | $ak:research-prompt ... | Native Skill discovery is supported; the Skill itself does not browse or call providers. |
Draft the research assignment
/ak:research-prompt Decide whether our approved onboarding guide should add a segment for first-time operations leaders. Use current customer evidence and primary market sources, separate fact from inference, and require source URLs for every finding/ak:research-prompt Decide whether our approved onboarding guide should add a segment for first-time operations leaders. Use current customer evidence and primary market sources, separate fact from inference, and require source URLs for every finding$ak:research-prompt Decide whether our approved onboarding guide should add a segment for first-time operations leaders. Use current customer evidence and primary market sources, separate fact from inference, and require source URLs for every findingThe argument is a research topic or decision. The output contract is fixed: exactly one paragraph, with no heading, preface, or second deliverable.
Follow the brief-building stages
- Gather context. Extract the decision, audience, deadline, known facts, constraints, and intended use from the conversation and relevant files.
- Make the brief self-contained. Explain the project and situation for a researcher with no prior context.
- Define one question. State the research question and the decision it will inform.
- Cover the decision. Add three to six inline numbered sub-questions without combining unrelated missions.
- Set evidence rules. Prefer official documentation, repositories, papers, filings, and changelogs. Treat forums and social posts as weak signals.
- Define contradiction and gap handling. Separate confirmed fact, inference, and unresolved uncertainty; revisit contradictions and claims with only one source.
- Specify the handoff. Require a source URL, specific claim, and one-line decision relevance for every finding, delivered as one detailed Markdown result by the eventual researcher.
Keep the brief free of hidden data and false certainty
The brief is not research evidence
Do not cite the generated paragraph as a finding. A later researcher must execute it, retrieve current sources, document conflicts, and produce the requested report. Review that report before changing messaging, publishing, contacting an audience, or spending budget.
Do not include secrets, private URLs, credentials, personal data, or unsupported facts in the assignment. Provider selection, network access, paid research tools, and sending project material outside the workspace are separate approvals for the execution stage.
Verify the output contract
A complete output should be:
- exactly one paragraph;
- understandable without prior conversation;
- centered on one research question and one decision;
- equipped with three to six inline numbered sub-questions;
- explicit about included and excluded scope, source hierarchy, contradiction handling, the gap round, and completion bar; and
- explicit about the per-finding URL, claim, and decision relevance required in the eventual detailed Markdown report.
The Skill normally returns the paragraph in the response. It does not create a file unless you separately request and approve a target path.
Troubleshoot and understand limits
| Symptom | Safe next step |
|---|---|
The runtime does not recognize ak:research-prompt | Confirm target and scope, restart the runtime, then follow Runtime cannot find a Skill or Agent. |
| The output has headings or multiple paragraphs | Ask for the fixed one-paragraph contract with no preface or second deliverable. |
| The brief contains several unrelated goals | Keep one decision and split the other missions into separate briefs. |
| Source requirements are vague | Require primary sources, per-finding URLs, contradiction handling, and a gap round. |
| The Skill starts researching | Stop and request only the assignment; execute it later with an approved research workflow. |
| The brief contains confidential context | Remove it before handoff and replace it with the minimum non-sensitive context required. |
The Skill cannot guarantee source availability, research quality, correctness, or a business outcome. It only improves the assignment contract.
Continue with the Marketing Kit overview or ak:docs after
the research result has been reviewed and accepted.
Understand the current ak:hub launcher limitation
Review the intended local Content Hub and Marketing Dashboard launcher without relying on stale paths or overstated features.
Turn stakeholder decisions into documents with ak:interview-docs
Run a guided interview that patches the user's own vision and decisions into a README, ADR, or structured project document.