agentleFS
VladTheProgrammerVladTheProgrammer/claude-asimov/.claude/skills/research-auto/SKILL.md

research-auto

Skill0 starsChanged 4 months agoMITView on GitHub

Autonomous (non-interactive) research for the autopilot. Produces docs/{task}/research.md like /research, but resolves open questions itself (wiki-ground -> answer -> default -> escalate) instead of a human conversation. Invoked by /autopilot; never run interactively.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 1

    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
  1. Distinctive:

    Before proposing structure or a design, CONSULT the captured, enforced system model:

    line 29

“Common” is measured: kept by at least 6 in 10 of the 3 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

You are running research in autonomous mode for the autopilot — no human to talk to. Produce the same artifact as /research (docs/{task}/research.md), in the same format, grounded in reality, with conclusions a second agent can re-verify.

Inputs (read first)

  • docs/{task}/ prior artifacts (facts.md, questions.md if present).
  • docs/objective.md (the human-owned frame) + the wiki (wiki/decisions/ = ratified authority; wiki/concepts/).
  • .claude/skills/research/SKILL.md defines the artifact format — match it exactly.

Autonomous decision procedure — wiki-ground -> answer -> default -> escalate

  1. Ground in the wiki first — read wiki/decisions/ + wiki/concepts/; do NOT re-research what's already decided (cite it).
  2. For each open question (from questions.md or the gap you're filling): find the answer from authoritative sources / the actual code; record the finding with a precise source (URL, file:line) so a verifier can re-check.
  3. Unresolved after a real search → choose the best-supported default; record it as provisional with rationale + alternative. Don't leave it dangling.
  4. Map the blast radius (required) — for the change this task implies, enumerate every other module / consumer / caller / serializer / UI / doc / test that depends on what will change. Record it as an explicit Cross-module impact section — a must-touch checklist that structure / better-plan must cover and review-auto later verifies nothing was left stale. The classic miss is changing the logic and leaving its UI/serializer behind.
  5. Escalate (write docs/{task}/ESCALATE.md) only if the answer would contradict the human-owned frame or is genuinely unknowable AND blocks progress.

Guardrails (every mode)

  • Ground in reality, never invent — every claim traces to a cited source, the wiki, or actual code.
  • Defaults, recorded — where /research would ask the human, pick the best-supported default and record what/why/alternative.
  • Escalate instead of guessing when it matters — frame-contradiction or unknowable-and-blocking → ESCALATE.md, not a half-artifact.
  • Self-contained, precisely-cited output at docs/{task}/research.md, matching the base skill's sections, including the required Cross-module impact / blast-radius section.

Architecture & design constraints (stewardship — read before deciding)

Before proposing structure or a design, CONSULT the captured, enforced system model:

  • wiki/architecture/system-map.md — the module catalog, dependency graph, the 5 enforced invariants, and the key seams (the core domain entry point, the adapter Protocol, the data feed, the shared record seam). The [tool.importlinter] contracts in pyproject.toml ENFORCE this — uv run lint-imports fails the build on a boundary cross. Extend the existing seams; do NOT re-architect an enforced boundary. If a task genuinely needs a boundary to move, that is an ADR + a contract change (say so), never a silent cross.
  • wiki/design-system/design-language.md (for any UI work) — the design tokens (theme.css), the component inventory, and the conventions (honest-n/a, the card/badge patterns, one-SPA tabs). Build UI in this language; do not invent ad-hoc styles or hardcode values that duplicate a token.

A design/structure that would violate a contract or the design language is wrong by construction.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.