From research-workspace
Reads `.research/` manifest files to produce a single orientation memo summarizing project research question, datasets, experiments, and open questions. Fast, read-only skill for getting context without scanning the whole repo.
How this skill is triggered — by the user, by Claude, or both
Slash command
/research-workspace:research-project-orienterThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Quickly orient an AI assistant inside a research workspace **without
Quickly orient an AI assistant inside a research workspace without
scanning the whole repository. Reads the .research/ manifest files
written by research-context-compressor and produces a single
in-conversation orientation memo.
Part of the research-hub skill pack; works alongside Zotero, Obsidian, and NotebookLM workflows but does not require any of them.
This skill is fast and read-only. If .research/ doesn't exist yet,
defer to research-context-compressor first.
Trigger phrases:
Not for:
research-context-compressor first.literature-triage-matrix.Read in this order:
.research/project_manifest.yml — top-level orientation. Required..research/experiment_matrix.yml — experiment status. Read if present..research/data_dictionary.yml — datasets. Read if present..research/decisions.md — recent ADRs. Read last 5 if present..research/open_questions.md — known unknowns. Read all..research/run_log.md — last 3 entries for context.Do not read source code, data files, or PDFs unless the manifest points you at a specific path AND the user's question requires it.
.research/ doesn't exist?Tell the user:
This project doesn't have a
.research/manifest yet. I can create one first (loadsresearch-context-compressorskill, takes ~30 seconds and writes 3 small YAML files), or I can fall back to scanning the repo directly (slower, more tokens). Which?
Don't auto-fall-back — ask first. If they pick "scan", read README.md +
docs/ + the top-level entrypoint, and produce the memo from that, but
caveat: "this orientation came from a one-shot scan; for more reliable
future sessions, run research-context-compressor once."
Single message in this exact structure:
## Project orientation: <project_name>
**Research question**: <one sentence from manifest>
**Stage**: <current_stage> · **Last updated**: <last_updated>
**Datasets** (<count>):
- `<name|id>`: <purpose|description; "(no description)" if both absent>
- ...
**Experiments** (<count>, by status):
- <status>: <id> — <hypothesis or method, one line; "(no hypothesis)" if both absent>
- ...
**Entrypoints**:
- `<path>` — <one-line purpose> # if list form (`main_entrypoints`)
- `<key>: <path>` # if object form (`entrypoints` map)
- ...
**Recent decisions** (<count>):
- <date>: <decision title>
**Open questions** (<count>):
- <question text>
- ...
**Evidence artifacts**:
- `<path>` — supports claim/figure
- ...
**Suggested next action**: <based on current_stage>
Length budget: ~200-400 tokens for typical project. Don't pad.
.research/, stop and tell the user the
manifest is incomplete — they should refresh with
research-context-compressor.paper-memory-builder does that.literature-triage-matrix does that.research_question is empty in the
manifest, say "no research question recorded; please add one to
.research/project_manifest.yml" rather than guessing.data/ or outputs/ unless directly asked.npx claudepluginhub wenyuchiou/ai-research-skills --plugin research-workspaceInspects a research repository and writes a compact .research/ workspace manifest (project_manifest.yml, experiment_matrix.yml, data_dictionary.yml) to save context for future AI sessions. Triggered by phrases like 'compress this project context'.
Documents codebases as-is by spawning parallel sub-agents to research files and synthesize findings into reports. Activates on /research or codebase understanding requests.
Performs preliminary codebase fact-finding and produces structured research reports. Use before cw-spec to understand unfamiliar codebases and generate enriched context.