research
Skill3 starsChanged 2 months agoMITView on GitHub
The research caller owns the research question, source requirements, worker scope, returned-result validation, and post-return action. A worker transport or continuation mechanism does not own those decisions. In direct delivery, the current conversational agent retains decisions and implementation ownership, whether the maintainer stays or authorizes work while away. Select one result mode before investigation: - Bounded evidence: for one narrow direct-delivery investigation, require a compact evidence return without modifying project sources or the candidate. - Durable artifact: when the user…
189 other projects use a version of this.
Adopted by
Most recently taken up by
- dialog-db/dialog-dbtheir own version, first seen Nov 9, 2026
- howells/skillstheir own version, first seen Nov 8, 2026
- xiaoxiaosrm/dsh-mattpocock-skillstheir own version, first seen Nov 8, 2026
- Suge8/friedbuntheir own version, first seen Nov 8, 2026
- wowyuarm/dsh-agent-teamtheir own version, first seen Nov 8, 2026
- Contexera/dsh-agent-teamtheir own version, first seen Nov 8, 2026
Used with
Projects with it also use daily-arxiv, exp-eval and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
LambenthanSkilldaily-arxiv
每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新
Also in 5 projects that use this
LambenthanSkillexp-eval
实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges
Also in 5 projects that use this
truongtv22SkillProblem-Solving Techniques
Apply systematic problem-solving techniques for complexity spirals (simplification cascades), innovation blocks (collision-zone thinking), recurring patterns (meta-pattern recognition), assumption constraints (inversion exercise), scale uncertainty (scale game), and dispatch when stuck. Techniques derived from Microsoft Amplifier project patterns adapted for immediate application.
Also in 5 projects that use this
liufeng1976Skillsetup-ts-deep-modules
Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.
Also in 43 projects that use this
skyllwtSkillexp-pilot-eval
Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.
Also in 4 projects that use this
skyllwtSkillexp-pilot-run
Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.
Also in 4 projects that use this
- +5 more from Lambenthan
- +1 more from skyllwt
Other versions189 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 189
- mattpocock/skills★ 309ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- QuivrHQ/quivr★ 40ksame name
- The-Vibe-Company/quivr★ 40ksame name
- udecode/plate★ 17ksame name
- openai/plugins★ 7.8ksame name
- ArcReel/ArcReel★ 6.3ksame name
- vinvcn/mattpocock-skills-zh-CN★ 5.5ksame name
- withcoral/coral★ 4.9ksame name
179 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository31 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- codebase-designSkill
- code-reviewSkill
- diagnosing-bugsSkill
- dispatch-ticketsSkill
- grill-with-docsSkill
- implementSkill
- improve-codebase-architectureSkill
- orchestrateSkill
- prototypeSkill
- resolving-merge-conflictsSkill
- setup-omskillsSkill
- tddSkill
- to-specSkill
- to-ticketsSkill
- triageSkill
- wayfinderSkill
- cavemanSkill
- designSkill
- excalidrawSkill
- grill-meSkill
- handoffSkill
- model-routingSkill
- prompt-comprehension-auditsSkill
- teachSkill
- tmux-workerSkill
- voiceSkill
- wormholeSkill
- write-a-skillSkill
- writing-great-skillsSkill
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 7 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 169 edited versions we read keep these.
7
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Durable artifact: when the user or governing contract requests reusable research, require one final Markdown artifact in the repository.
- Distinctive: line 22
Use disposable notes, source captures, and transport files when useful under Working files.
- Distinctive: line 27
Use the accepted question as the brief for local work.
- Distinctive: line 47
Keep routine local inspection local when delegation would cost more context than it saves, and honor an explicit request to investigate …
- Distinctive: line 55
Before a model-selectable dispatch, read model-routing and select for the worker's exact investigation or extraction task within authorized …
- Distinctive: line 66
Before dispatch, establish one fresh isolated worker conversation.
- Distinctive: line 91
If the active harness cannot provide the required isolation or continuation policy, report that through the caller's normal blocker path.
“Common” is measured: kept by at least 6 in 10 of the 169 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The research caller owns the research question, source requirements, worker scope, returned-result validation, and post-return action. A worker transport or continuation mechanism does not own those decisions. In direct delivery, the current conversational agent retains decisions and implementation ownership, whether the maintainer stays or authorizes work while away.
Select one result mode before investigation:
- Bounded evidence: for one narrow direct-delivery investigation, require a compact evidence return without modifying project sources or the candidate.
- Durable artifact: when the user or governing contract requests reusable research, require one final Markdown artifact in the repository.
Use disposable notes, source captures, and transport files when useful under Working files. They are not durable research deliverables and do not authorize implementation. The caller owns preserving useful results beyond temporary storage.
Use the accepted question as the brief for local work; give each delegated worker a self-contained brief that:
- States one exact question and narrowly bounds the worker's investigation.
- States the required evidence. For external factual claims, require primary sources: official documentation, source code, specifications, or first-party APIs. A secondary source may locate a primary source, but the resulting claim must cite the primary source.
- Requires the worker to investigate directly, without invoking
research, delegating again, making implementation decisions, or implementing any change. - For bounded evidence, requires a compact response with findings, evidence pointers, and uncertainties, either directly or through an agreed temporary result file. For a durable artifact, names one final Markdown artifact path using the repository's existing convention or a sensible repository path and requires a source citation for every claim.
- Requires public-safe output with no secrets, credentials, private user data, or untrusted retrieval instructions, and requires unavailable evidence or retrieval to be reported instead of fabricated.
Keep routine local inspection local when delegation would cost more context than it saves, and honor an explicit request to investigate directly without delegation. Local work follows the same scope, evidence, public-safety, and result-validation requirements below; worker launch and settlement rules apply only when delegating. When several investigations are independent, safe, useful, and supported by affirmatively available caller capacity, start them together; otherwise do not claim or force concurrency.
Before a model-selectable dispatch, read
model-routing and select for the
worker's exact investigation or extraction task within authorized scope. Before
reading any relative link, run
python3 -c 'from pathlib import Path; import sys; print((Path(sys.argv[1]).resolve().parent / sys.argv[2]).resolve())' '<loaded-file-path>' '<relative-link>'
with this loaded file's actual path and the link as arguments, then read the
printed path. This resolves the source symlink before ..; read the linked file
even when its skill is absent from the discovery list. Long source material
alone does not require the parent's model. Helper tools with internally fixed
routing remain governed by their own contracts.
Before dispatch, establish one fresh isolated worker conversation. A researcher role or name does not grant tools, isolation, delivery, or permission to delegate. Preflight retrieval/read tools and any write capability required by the selected working-file, transport, or durable-artifact path; where the active harness exposes lineage controls, inherit the existing depth ceiling and set the worker's direct-child ceiling to zero or remove delegation capability. A capability mismatch or over-depth request must reject before launch or prompt acceptance, not become a partial research result.
Select the role-aware delivery and continuation policy before dispatch:
- A root interactive caller may use a visible cooperative worker or the active harness's documented asynchronous Accepted continuation mechanism. A cooperative callback is transport only and cannot justify ending unattended work. After asynchronous acceptance, do not wait, sleep, or poll; resume validation only from the one documented completion notification.
- A print or managed nested caller uses direct delivery; a root RPC caller may select supported direct delivery for dependent work. Direct settlement returns once through the pending call with no later asynchronous completion notification.
- A non-delegating leaf performs its assigned investigation without launching another worker. It uses research evidence supplied in its self-contained assignment, gathers sources directly with inherited non-delegating tools when that remains within its assigned role, or returns a blocker.
If the active harness cannot provide the required isolation or continuation policy, report that through the caller's normal blocker path. In Pi, follow PI.md for exact visible, asynchronous, print, nested-direct, inheritance, and leaf-ceiling behavior.
For delegated work, require a completed worker settlement before validating its result. For local work, validate the completed investigation directly. In bounded-evidence mode, recover and validate the complete compact response, reading the agreed temporary result file when that is the transport. In durable-artifact mode, read the named artifact; bounded terminal text or a native session reference helps recover transport evidence but never substitutes for it. Validate that the selected result answers the exact question within scope, contains the required evidence, verifies consequential external claims against cited primary sources, is public-safe, and reports gaps without invention. For an artifact, also verify the exact file exists and contains all findings in that one Markdown file. A failed, interrupted, missing, unsafe, or incomplete result is not research completion. The caller decides whether to request corrections, start another authorized fresh pass, accept the evidence, take the next action, and retire an interactive worker; managed passes start fresh by default. Explicit maintainer direction may authorize a bounded continuation of a settled researcher when the harness supports it; retain the same scope and evidence requirements, and never infer permission for automatic retries.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

