Agent-Skills.md

Agent Skills: Source Manifest

Build or validate the source manifest for the source-tutorial pipeline from user-provided URLs/files.

UncategorizedID: willoscar/research-units-pipeline-skills/source-manifest
51039

Install this agent skill to your local

pnpm dlx add-skill https://github.com/WILLOSCAR/research-units-pipeline-skills/tree/HEAD/.codex/skills/source-manifest

Skill Files

Browse the full folder contents for source-manifest.

Download Skill

Loading file tree…

.codex/skills/source-manifest/SKILL.md

Skill Metadata

Name
source-manifest
Description
"Build or validate the source manifest for the source-tutorial pipeline from user-provided URLs/files."

Source Manifest

Triggers & routing

  • Trigger: source manifest, sources list, tutorial sources, url list, 资料清单, 教程来源.
  • Use when: source-tutorial 的 C1，需要把多源输入落成统一的 sources/manifest.yml，并在内容不完整时显式阻塞。
  • Skip if: 已经有完整且经过确认的 sources/manifest.yml。
  • Network: none.
  • Guardrail: 不要伪造来源；manifest 不完整时应返回 BLOCKED，而不是假装完成。

Goal: collect the source candidates for source-tutorial into one explicit manifest before ingest starts.

Inputs

  • GOAL.md
  • DECISIONS.md

Outputs

  • sources/manifest.yml

Workflow

  1. Read GOAL.md and any existing DECISIONS.md notes to understand what kinds of sources the tutorial is expected to use.
  2. If sources/manifest.yml does not exist, scaffold it with a minimal example.
  3. Validate that each source has:
    • source_id
    • kind
    • locator
    • label
    • for kind: video, prefer transcript_locator unless the platform exposes public subtitles
  4. Reject placeholder/example-only manifests.
  5. Return success only when at least one real source exists.

Accepted kind

  • webpage
  • pdf
  • markdown
  • repo
  • docs_site
  • video

For video:

  • preferred: add transcript_locator
  • Bilibili may succeed from public subtitles when available
  • YouTube should currently be paired with transcript_locator

Script

Quick Start

  • uv run python .codex/skills/source-manifest/scripts/run.py --workspace <workspace>

All Options

  • --workspace <dir> (required)
  • --unit-id <U###>
  • --inputs <semicolon-separated>
  • --outputs <semicolon-separated>
  • --checkpoint <C#>

Examples

  • Scaffold or validate the manifest:
    • uv run python .codex/skills/source-manifest/scripts/run.py --workspace <workspace>

Troubleshooting

Issue: the unit keeps blocking after scaffold

Cause:

  • The manifest still contains example placeholders.

Fix:

  • Replace the scaffold entry with real URLs or local file paths, then rerun the unit.