fast-track-ai-research

1. Read fast-track-ai/INDEX.md first. State the research question and the acceptance boundary before gathering evidence.
2. Prefer primary sources. Retrieve and read the source itself rather than relying on search-result snippets. Use more than one independent source for a material conclusion when available.
3. Tie every material claim to a direct citation. Distinguish observed facts from inference. Label claims that depend on one source or potentially stale evidence.
4. Report credible disagreement under Conflicts. Report inaccessible material, freshness concerns, single-source dependence, inference, and other uncertainty under Limitations. When the retrieved evidence does not support an answer, state in the conclusion that the result is inconclusive and name the evidence that would settle it, instead of forcing an answer.
5. Write exactly one result to fast-track-ai/research/<slug>.md from ./assets/research.md. Use a descriptive kebab-case slug. Frontmatter contains captured and owner only; source records belong in the body so the artifact can cite multiple sources. Do not add a lifecycle status or singular source field. When fast-track-ai/research/<slug>.md already exists, rewrite it only when the user asked to refresh that research; otherwise choose a distinct slug. Never overwrite an existing research file silently.
6. Run ../fast-track-ai-init/scripts/fast-track-ai-index.sh <target> using the owner copy and report any validation error.
7. When invoked by another skill as an auxiliary step, write the artifact, regenerate the index, return conclusion, limitations, and artifact path to the caller, and present nothing to the user. A direct call reports the result, path, and index outcome without a . line.

Path resolution: ./ paths and ../fast-track-ai-* sibling paths resolve from this skill's installed directory, and fast-track-ai/ paths resolve from the target repository root.