Skill

researcher

Generates or refreshes research artifacts and RLM evidence for tickets in research stage of feature development workflow. Validates RLM outputs like *-rlm-targets.json.

Python

Bash

automation

developer-tools

npx claudepluginhub grinrus/ai_driven_dev --plugin feature-dev-aidd

Configuration

Model: inherit

Tool Access

This skill is limited to using the following tools:

ReadEditWriteGlobBash(rg *)Bash(sed *)Bash(python3 ${CLAUDE_PLUGIN_ROOT}/skills/researcher/runtime/research.py *)Bash(python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-flow-state/runtime/set_active_feature.py *)Bash(python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-flow-state/runtime/set_active_stage.py *)Bash(python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-flow-state/runtime/tasks_derive.py *)

Preview

Follow `feature-dev-aidd:aidd-core`.

Supporting Assets

runtime/research.pytemplates/research.template.md

SKILL.md

Similar Skills

research

Executes multi-pass research on topics for sprints, adding 3-5 evidence-based claims per pass to claims.json using web search, DeepWiki for GitHub repos, and file tools.

grainulator

research

136

Orchestrates research workflows from question definition to evidence-based findings documentation for technical, requirements, literature, and codebase topics.

3 files

maister-copilot

research-plan-implement

Orchestrates end-to-end Research-Plan-Implement pipeline using parallel subagents per phase with file artifacts for communication. Use for complex features needing codebase research, planning, and implementation in one session.

rpikit

Stats

Stars5

Forks0

Last CommitApr 17, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

Follow feature-dev-aidd:aidd-core.

Steps

Set active feature and stage research.
Run python3 ${CLAUDE_PLUGIN_ROOT}/skills/researcher/runtime/research.py --ticket <ticket> --auto.
Re-run the same entrypoint with optional overrides (--paths, --keywords, --note) when targeted refresh is needed.
Validate RLM outputs (*-rlm-targets.json, *-rlm-manifest.json, *-rlm.worklist.pack.json, optional *-rlm.pack.json).
Run subagent feature-dev-aidd:researcher. First action: read RLM pack/worklist.
In --auto mode run bounded canonical finalize recovery once (rlm_finalize --bootstrap-if-missing) from stage runtime.
If RLM is still pending after auto recovery, return deterministic pending reason + explicit next action (python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-rlm/runtime/rlm_finalize.py --ticket <ticket>) and append research handoff via python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-flow-state/runtime/tasks_derive.py --source research --append.
Ready path: return /feature-dev-aidd:plan-new <ticket> only when RLM readiness is ready and the research document is synchronized with the current evidence.
Pending or blocked path: if readiness is not ready, return the deterministic blocker or pending output instead of guessing a downstream handoff.
Runtime-path safety: execute only canonical runtime commands from this contract (python3 ${CLAUDE_PLUGIN_ROOT}/skills/.../runtime/...).
Root-relative /skills/... runtime paths are forbidden; use only ${CLAUDE_PLUGIN_ROOT}/skills/.../runtime/....

Command contracts

`python3 ${CLAUDE_PLUGIN_ROOT}/skills/researcher/runtime/research.py`

When to run: always as canonical researcher pipeline entrypoint.
Inputs: --ticket <ticket> with optional path/keyword/note overrides.
Outputs: research artifacts, stage status, and RLM readiness/handoff markers (rlm_status, rlm_pack_status).
Failure mode: non-zero exit when required inputs/artifacts are missing or RLM artifact generation fails.
Next action: fix missing inputs/paths, rerun this entrypoint; in auto mode finalize/bridge runs once, then pending outcomes include deterministic reason_code + next_action.

Notes

Planning stage: AIDD:ACTIONS_LOG: n/a.

Migration policy

Legacy pre-RLM research context/targets artifacts are ignored by runtime/gates.
For old workspace state, rerun: python3 ${CLAUDE_PLUGIN_ROOT}/skills/researcher/runtime/research.py --ticket <ticket> --auto.
If rlm_status remains pending after bounded auto recovery, hand off to shared owner: python3 ${CLAUDE_PLUGIN_ROOT}/skills/aidd-rlm/runtime/rlm_finalize.py --ticket <ticket>.

Additional resources

Research template source: templates/research.template.md (when: drafting or reviewing aidd/docs/research/<ticket>.md; why: keep artifact shape aligned with canonical workspace template).
Shared RLM owner skill: ../aidd-rlm/SKILL.md (when: rlm_status is not ready or RLM pack is missing; why: run canonical shared RLM API outside stage-local orchestration).