research
This skill uses Claude hooks which can execute code automatically in response to events. Review carefully before installing.
MANDATORY: Follow the complete shared lifecycle contract in ${CLAUDE_PLUGIN_ROOT}/skills/shared/run-lifecycle.md with declared skill research.
Research Skill
MANDATORY: READ ENTIRE FILE before composing user-facing prose: ${CLAUDE_PLUGIN_ROOT}/skills/shared/readability-style.md.
Collaborative best-effort read-only-repo research task with a fixed-shape lane topology. The research phase runs a planner pre-pass that decomposes RESEARCH_QUESTION into 2–4 focused subquestions, then four Codex-first lanes (architecture / edge cases / external comparisons / security) covering those subquestions, each with a per-lane Claude Agent fallback when Codex is unavailable or fails. The validation phase runs three reviewers in parallel: 1 Claude Code Reviewer subagent + 1 Codex + 1 Cursor. Produces a structured research report; tracked repo files are not modified by Claude's Edit | Write | NotebookEdit tool surface while the fresh research-* activation sentinel exists (mechanically enforced by the skill-scoped PreToolUse hook permitting only canonical /tmp and the larch session cache), while Bash and the external Cursor/Codex reviewers run with full filesystem access and are prompt-enforced only: see the Read-only-repo contract below. May invoke /issue via the Skill tool to file research-result issues.
Anti-halt continuation reminder. After every child Skill tool call (e.g., /issue) returns, IMMEDIATELY continue with this skill's NEXT numbered step: do NOT end the turn on the child's cleanup output, and do NOT write a summary, handoff, status recap, or "returning to parent" message: those are halts in disguise. The rule is strictly subordinate to any explicit non-sequential control-flow directive in THIS file (e.g., skip to Step N, bail to cleanup, jump back, loop back, fall through, break out). A normal sequential proceed to Step N+1 instruction is the default continuation this rule reinforces, NOT an exception. → shared/subskill-invocation.md#anti-halt
Flags: Parse flags from the start of $ARGUMENTS before treating the remainder as the research question. Flags may appear in any order; stop at the first non-flag token. After stripping all flags, save the remainder as RESEARCH_QUESTION.
--no-issue(boolean): Set a mental flagRESEARCH_NO_ISSUE=true. When set, Step 3.5 (auto-issue) is skipped: no GitHub issue is created for the research results. Default:RESEARCH_NO_ISSUE=false. WhenRESEARCH_NO_ISSUE=false(default), Step 3.5 creates a GitHub issue containing the research question, full report, and token spend metadata via/issuesingle mode.--run-id <ID>: Use${CLAUDE_PLUGIN_ROOT}/skills/shared/run-id-flag.mdfor the shared contract.
Fail-closed unknown-flag guard: After flag parsing finishes, inspect the next token of $ARGUMENTS. If it begins with --, abort: