AGENT SKILL

finding-outputter

Output contract for a research finding. Owns the finding's frontmatter, sections, tier-tagged sources, and write sink. Injected into the researcher by the research-topic conduct verb — not invoked directly.

License 许可
MIT
最近更新
2026-11-12
Stars
★ 34
Forks
2

01 前言 Front Matter

name
finding-outputter
description
Output contract for a research finding. Owns the finding's frontmatter, sections, tier-tagged sources, and write sink. Injected into the researcher by the research-topic conduct verb — not invoked directly.
user-invocable
false
disable-model-invocation
true

03 SKILL.md 正文

Finding Outputter

You write one finding for a single focus area. This document owns the finding's shape end to end: what it is, where it goes, and what it must never do.

Sink

Write the finding to the single output path injected into your task prompt, and to no other path.

Injected Values

Your task prompt injects everything the frontmatter needs — do not invent or derive it:

  • finding template — the resolved topic-research-finding template to render.
  • output path — where to write the finding.
  • round — the round number this finding belongs to.
  • depth — the resolved depth the finding was researched or composed under.
  • question — the focus area's question.
  • source profile — the profile this finding was researched through.
  • timestamp — the ISO datetime for date and last_updated.
  • author — the value for author and last_updated_by.

Shape

The topic-research-finding template is the finding's shape — frontmatter and body skeleton both — and is injected into your task prompt. Fill its placeholders from the injected values. If no template was injected, refuse rather than improvise a shape.

Fill the template from the injected values, observing the deltas it cannot express on its own:

  • status is always complete.
  • source_profile is always the injected source profile, and question the injected question byte for byte, each overriding any value the template carries.
  • Omit the revision/repository pair entirely — a finding is not code-state-anchored.
  • The typed-linkage slots (parent, relates_to) are omit-when-empty: emit a key only when it carries a value. Never write parent: "" or relates_to: [] — the validator rejects those literal empties.

Body

The template you loaded is authoritative for the finding's sections — follow its structure. The guidance below is about the content to write, not a fixed set of headings: where the template has a matching section, fill it out this way, and use your judgement for any section the template shapes differently.

  • Question — the focus area's question, restated.

  • Findings — standalone research prose that answers the question. It reads on its own: no round narration, no "in this round", no changelog of what you did. A reader who opens only this finding must understand it.

  • Sources — one bullet per source, each carrying its reputation tier and recorded domain or venue, and the source's metadata block when it has one (see Sources Metadata Block):

    - [Title](url) — tier-1 — {source domain/venue}
      - doi: 10.7717/peerj.4375
      - openalex_id: W2741809807
      - published: 2018-02-13
      - accessed: 2026-10-24
    

    When the profile's sources held nothing relevant, Sources reads None found. instead.

Tier every source tier-1/tier-2/tier-3 by the source profile's vocabulary, from venue identity alone. The tier text carries any suffix the profile prescribes, such as tier-3 (retracted).

A source cited at several tiers lists the others in an also suffix, after any retracted or withdrawn mark: tier-3 (also tier-2, tier-1), tier-3 (retracted; withdrawn; also tier-1).

Sources Metadata Block

A Sources entry may carry a metadata block: a nested - key: value list directly under its line. The block is optional, and an entry without one is valid. A researcher writes each block from the CLI's output, under the rules below.

  • Block lines are the CLI's metadata entries for the cited source, copied unchanged and in emitted order. Never write a block line from anywhere else: not from a page, an abstract, a record's other fields, or memory.
  • Keys are lowercase snake_case, values are single-line, and a key with no value is omitted.
  • url, title and venue never appear in the block: they stay on the entry line.
  • The CLI emits only recommended keys, in table order. Copying in emitted order keeps that order.
  • The template's - {source-metadata} slot line repeats under every Sources entry. Under each, it becomes one - key: value line per entry of that source's metadata map, in emitted order. When the source has no map, delete the whole slot line: never leave it as written, and never fill it from anything else.

A composer writes each block from the notes' blocks instead, under the citing-once rule. Every line still comes from a block a note carries.

The recommended keys, in order:

Key Group Value form
doi identifier bare and lowercase: 10.7717/peerj.4375
arxiv_id identifier bare, with the version retrieved: 2305.12345v2
openalex_id identifier the bare W ID: W2741809807
pmid identifier
issn identifier
published date ISO 8601 at the precision the source declares: 2023, 2023-05 or 2023-05-12
updated date as published
accessed date the UTC date of the CLI call the map came from: 2026-10-24
type bibliographic
authors bibliographic
publisher bibliographic
volume bibliographic
issue bibliographic
pages bibliographic
license bibliographic
journal_ref source-specific
primary_category source-specific
oa_status source-specific
pdf_url source-specific

The CLI fills the core keys, those with a value form above, wherever the source provides them. journal_ref is a citation string, not a venue, so it does not duplicate the entry line.

04 同仓库更多 Skills

atomicinnovation
atomicinnovation/accelerator

explore-pr

Build an interactive, design-first and picture-first exploration page for a pull request — its architecture, abstractions and design decisions drawn as diagrams, a guided tour in reading order with code as the evidence, an attention map of where the risk sits, a ledger of its riskiest operations, prior review findings re-checked against the code, and simulations specific to the PR — with the reviewer's verdicts and notes saved. Use when the user wants to explore, understand or review a large pull request or its design, asks for a review page, walkthrough or visual explanation of a PR, or is struggling to review a big change.

设计师视觉设计
★ 34
atomicinnovation
atomicinnovation/accelerator

update-work-item

Update fields (status, priority, tags, parent, etc.) of an existing work item. Use to transition status, change priority, manage tags, or edit any frontmatter field. No transition enforcement — arbitrary changes are allowed.

★ 34
atomicinnovation
atomicinnovation/accelerator

sync-work-items

Reconcile local work items in meta/work/ with the active remote tracker named by work.integration. Use when the user wants to sync, push, or pull work items to or from Jira or Linear, preview what a sync would change, or reconcile divergent local and remote state.

★ 34
atomicinnovation
atomicinnovation/accelerator

sync-work-items-recovery

Recovery procedures for work sync failures that need a person — terminal rows, refused pending pushes and refused orders. Read by sync-work-items — not invoked directly.

★ 34
atomicinnovation
atomicinnovation/accelerator

stress-test-work-item

Interactively stress-test a work item by grilling the user on scope, assumptions, acceptance criteria, edge cases, and dependencies to surface issues, gaps, and flawed assumptions before implementation is planned.

测试与质量
★ 34
atomicinnovation
atomicinnovation/accelerator

review-work-item

Review a work item through multiple quality lenses and collaboratively iterate based on findings. Use when the user wants to evaluate a work item before implementation or escalation.

★ 34