Skip to content
Back to search

content-to-skill

Convert source material into executable Agent Skill packages

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/gnipbao--content-to-skill/badge.svg)](/listings/gnipbao--content-to-skill)

HTML

<a href="/listings/gnipbao--content-to-skill"><img src="/api/listings/gnipbao--content-to-skill/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Copy this folder into any skills-compatible runtime's skill directory, or install it through your runtime's skill installer if one is available.

Common local paths:

$CODEX_HOME/skills/content-to-skill
~/.codex/skills/content-to-skill
$CLAUDE_HOME/skills/<skill-name>
~/.claude/skills/<skill-name>

For this repository, keep the folder name `content-to-skill`.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 279 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. ---
  2. name: content-to-skill
  3. description: Converts arbitrary source material into an executable structured Codex skill package. Use when the user provides text, notes, documents, PDFs, transcripts, videos, courses, articles, repositories, workflows, chat logs, or rough ideas and asks to turn them into a Skill, SKILL.md, skill scaffold, reusable agent workflow, operational prompt, or structured capability. Also use for 把内容变成skill, 文档转skill, 视频转skill, 课程转skill, 工作流沉淀成skill, or 从资料生成可执行Skill.
  4. ---
  5. # Content To Skill
  6. Content To Skill turns messy source material into a runnable skill package. It extracts repeatable mechanisms from content, rejects unsupported claims, and produces a `SKILL.md` plus any references, examples, scripts, and validation prompts needed for another agent to execute the method later.
  7. Default to Chinese unless the user asks for another language. Prefer creating or patching files when a writable target path is available.
  8. ## Resource Guide
  9. - Read `references/source-intake.md` before processing documents, transcripts, videos, multiple files, or unclear source boundaries.
  10. - Read `references/skill-construction-protocol.md` before writing a `SKILL.md`, scaffold, references, examples, or validation plan.
  11. - Read `references/evaluation-rubric.md` when reviewing a generated skill, deciding whether it is publishable, or repairing a weak conversion.
  12. - Run `scripts/check_skill_package.py /path/to/generated-skill` after creating or materially editing a local skill package.
  13. - Read `examples/retest-prompts.md` when you need realistic forward tests for a newly generated skill.
  14. ## First Principle
  15. A source is not the skill. A skill is the compressed control system that lets a future agent repeat the useful behavior without rereading the whole source.
  16. Every conversion must preserve this chain:
  17. ```txt
  18. source evidence -> reusable mechanism -> executable workflow -> output contract -> validation signal
  19. ```
  20. ## Core Workflow
  21. ### 1. Route The Request
  22. 模式选择: choose the smallest mode that preserves the source evidence and produces an executable artifact.
  23. Choose the smallest mode that fits:
  24. | Trigger | Mode | Required action |
  25. |---|---|---|
  26. | User pastes short text or a clear method | QUICK_CONVERT | Extract mechanisms and produce a first-pass `SKILL.md` or patch. |
  27. | User provides documents, folders, transcripts, video, or many sources | SOURCE_AUDIT | Build a source inventory, evidence boundary, mechanism cards, then generate. |
  28. | User asks for files, scaffold, installable skill, or gives a path | PACKAGE_BUILD | Create files first, then validate. |
  29. | User provides an existing generated skill and says it failed | REPAIR_FROM_FEEDBACK | Compare source intent, generated behavior, and failure signal before patching. |
  30. | Source cannot be accessed | BLOCKED_SOURCE | Ask for the missing file, transcript, text, or accessible path; do not invent source content. |
  31. If multiple modes match, prefer the one with stronger evidence requirements: REPAIR_FROM_FEEDBACK, PACKAGE_BUILD, SOURCE_AUDIT, QUICK_CONVERT.
  32. ### 1.5. CHECKPOINT / STOP Gates
  33. Emit `CHECKPOINT / STOP` and wait before:
  34. - overwriting an existing skill folder with unrelated local edits
  35. - publishing or pushing a generated skill to a public repository
  36. - processing private, paid, sensitive, or inaccessible source material
  37. - expanding one source into a router, skill family, or SkillBank
  38. - claiming `full_test` quality without actually running an independent or fresh-context test
  39. The checkpoint must name the target path or repository, files affected, validation command, rollback point, and exact approval needed.
  40. ### 2. Establish Evidence Boundary
  41. Identify what was actually read and what remains unavailable.
  42. Capture:
  43. - source type: text, document, PDF, spreadsheet, slide deck, audio, video, repository, web page, notes, or mixed
  44. - accessible artifacts: pasted text, local paths, transcripts, extracted text, screenshots, metadata, or summaries
  45. - unavailable artifacts: video without transcript, unreadable binary files, private links, missing attachments, or unsupported formats
  46. - user goal: new skill, skill family, prompt workflow, repair, or evaluation
  47. - target runtime: Codex skill folder, single `SKILL.md`, repository scaffold, or patch plan
  48. For video or audio, first obtain a transcript, chapter notes, subtitles, or user-provided summary using available tools. If none is available, ask for the transcript or permission to extract it. Never infer detailed methods from a title alone.
  49. ### 3. Extract Mechanism Cards
  50. Do not summarize the source paragraph by paragraph. Extract portable controls:
  51. ```md
  52. ## Mechanism Card
  53. Name:
  54. Source evidence:
  55. Trigger:
  56. User job:
  57. Decision rule:
  58. Procedure:
  59. Output:
  60. Quality signal:
  61. Failure mode:
  62. Skill location:
  63. Keep / merge / discard:
  64. ```
  65. Keep only mechanisms that change future agent behavior. Discard anecdotes, slogans, duplicate examples, and unsupported claims unless they become examples, boundaries, or tests.
  66. ### 4. Choose Skill Shape
  67. Select one production shape before writing files:
  68. - Single procedural skill: one bounded repeated task.
  69. - Engineering workflow pack: coding, debugging, architecture, tests, issues, or PR workflows.
  70. - Methodology toolbox: router plus focused diagnostic skills.
  71. - Cognitive distillation skill: a person, worldview, creator, company, or field turned into a perspective lens.
  72. - Skill family or SkillBank: many independent skills with routing, acquisition, versioning, or state.
  73. Default to a single procedural skill unless the source clearly requires routing, shared state, multiple domains, or independent subskills.
  74. ### 5. Generate The Skill Package
  75. When a writable path is available and the user asks to create or convert, create files before the final answer.
  76. Minimum package:
  77. ```txt
  78. skill-name/
  79. ├── SKILL.md
  80. ├── agents/
  81. │ └── openai.yaml
  82. └── examples/
  83. └── retest-prompts.md
  84. ```
  85. Add `references/` when the source has detailed protocols, rubrics, evidence notes, examples, or domain facts that should not bloat `SKILL.md`.
  86. Add `scripts/` only when repeated mechanical checks, extraction, conversion, or validation should be deterministic.
  87. For public GitHub repositories, add only the human-facing files needed for reuse:
  88. ```txt
  89. README.md
  90. LICENSE
  91. .gitignore
  92. test-prompts.json
  93. ```
  94. Keep the skill package itself portable: `SKILL.md`, `agents/`, `references/`, `examples/`, and `scripts/` must still work if copied into any skills-compatible runtime.
  95. Generated `SKILL.md` must include:
  96. - frontmatter with only `name` and `description`
  97. - overview and first principle
  98. - resource guide with direct links to references and scripts
  99. - routing or operating modes
  100. - stable workflow that produces decisions or artifacts
  101. - output protocols
  102. - boundaries and anti-patterns
  103. - quality standard and validation loop
  104. ### 6. Validate
  105. For local packages:
  106. 1. Run the repository or generated package validator when available.
  107. 2. Run `scripts/check_skill_package.py /path/to/generated-skill`.
  108. 3. Read validator failures, patch once, and rerun.
  109. 4. If validation still fails, report the failing command, likely cause, and rollback point.
  110. For non-local outputs, use the checklist in `references/evaluation-rubric.md` and label the result as dry-run.
  111. ### 6.5. Failure Branches
  112. | Trigger | Required action | Do not do |
  113. |---|---|---|
  114. | Source cannot be read | Ask for accessible source, transcript, or a local path | Do not invent source mechanisms |
  115. | Target folder already exists | Inspect files and patch only intended paths | Do not overwrite unrelated files |
  116. | Existing skill fails validation | Patch one minimal structural issue and rerun | Do not add broad new sections to chase score |
  117. | Public repo requested | Validate locally, create README/LICENSE, then ask or use explicit approval before publishing | Do not push unvalidated files |
  118. | User asks for a skill family | Create a mechanism map and stop at CHECKPOINT / STOP | Do not explode one source into many skills by default |
  119. | Test tooling is unavailable | Mark result `dry-run` and provide retest prompts | Do not call it `full_test` |
  120. ### 7. Report
  121. Final reports should include:
  122. ```md
  123. ## Skill Created
  124. Source boundary:
  125. Generated path:
  126. Skill shape:
  127. Files changed:
  128. Validation:
  129. Rollback point:
  130. Next retest:
  131. ```
  132. Keep design explanation short unless the user explicitly asks for the full extraction notes.
  133. ## Output Protocols
  134. ### Source-To-Skill Brief
  135. ```md
  136. ## Source-To-Skill Brief
  137. Source boundary:
  138. Root job:
  139. Target user:
  140. Mechanisms kept:
  141. Mechanisms discarded:
  142. Skill shape:
  143. Package plan:
  144. Validation plan:
  145. Open uncertainty:
  146. ```
  147. ### Mechanism Extraction Report
  148. ```md
  149. ## Mechanism Extraction
  150. Evidence read:
  151. High-signal mechanisms:
  152. Conflicts or weak evidence:
  153. Candidate skill modules:
  154. Discarded material:
  155. Recommended package:
  156. ```
  157. ### Generated Package Report
  158. ```md
  159. ## Skill Created
  160. Path:
  161. Name:
  162. Description trigger:
  163. Source boundary:
  164. Files created or changed:
  165. Validation commands:
  166. Validation result:
  167. Retest prompt:
  168. Rollback point:
  169. ```
  170. ## Boundaries
  171. - Do not claim to have read inaccessible documents, videos, private links, or repositories.
  172. - Do not turn every idea in the source into a rule. Prefer fewer mechanisms that can be executed.
  173. - Do not copy large passages from source material into the skill. Extract controls, examples, and evidence notes.
  174. - Do not create a skill family when one procedural skill will work.
  175. - Do not create scripts for tasks that are better handled by clear instructions.
  176. - Do not add README, install guides, or public docs unless the user asks for a public repository.
  177. - Do not overwrite user files outside the target skill folder without explicit approval.
  178. - Do not claim benchmark quality without validation or forward tests.
  179. ## Anti-Patterns
  180. - Summary-as-skill: the output explains the source but cannot guide future behavior.
  181. - Template flood: the skill has many sections but no decision rules.
  182. - Evidence laundering: weak or unavailable source material becomes confident instruction.
  183. - Skill bloat: every chapter becomes a subskill.
  184. - Tool fantasy: the skill assumes unavailable transcription, parsing, browser, or API tools.
  185. - No retest: the conversion is never checked against a realistic future request.
  186. ## Quality Standard
  187. A successful conversion:
  188. - names the source boundary honestly
  189. - extracts mechanisms, not just themes
  190. - has a clear trigger description
  191. - can run without the original source in context
  192. - has stable workflow steps and output contracts
  193. - stores detailed evidence in references, not the main prompt
  194. - names failure branches and stop conditions
  195. - includes retest prompts or fixtures
  196. - passes structural validation when files exist locally
  197. - stops at CHECKPOINT / STOP before public publishing, broad generation, or sensitive source processing
  198. Default score:
  199. ```md
  200. Root job clarity: 0-20
  201. Mechanism extraction: 0-20
  202. Executable workflow: 0-20
  203. Output usability: 0-15
  204. Evidence honesty: 0-10
  205. Validation loop: 0-10
  206. Concise expression: 0-5
  207. Total: 100
  208. ```
  209. Below 70: do not publish. 70-84: usable MVP. 85-94: publishable. 95+: benchmark-quality after independent or forward-test evidence.
README.md · README · 88 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Content To Skill
  2. Content To Skill converts source material into a runnable Agent Skill package. It is useful when you have notes, documents, transcripts, videos, course material, workflows, repositories, or rough methods and want to turn them into a repeatable `SKILL.md` with references, examples, scripts, and validation prompts.
  3. The core rule is simple:
  4. ```txt
  5. source evidence -> reusable mechanism -> executable workflow -> output contract -> validation signal
  6. ```
  7. ## When To Use
  8. Use this skill when you want to:
  9. - convert documents, notes, transcripts, or workflow memos into a structured skill
  10. - package a repeatable method as `SKILL.md`
  11. - audit a weak generated skill and repair it from source evidence
  12. - create a public or private skill repository with validation prompts
  13. - preserve the useful mechanism of source material without copying the source wholesale
  14. ## Install
  15. Copy this folder into any skills-compatible runtime's skill directory, or install it through your runtime's skill installer if one is available.
  16. Common local paths:
  17. ```txt
  18. $CODEX_HOME/skills/content-to-skill
  19. ~/.codex/skills/content-to-skill
  20. $CLAUDE_HOME/skills/<skill-name>
  21. ~/.claude/skills/<skill-name>
  22. ```
  23. For this repository, keep the folder name `content-to-skill`.
  24. ## Usage
  25. Invoke it by name:
  26. ```text
  27. Use $content-to-skill to convert this transcript into a runnable skill package.
  28. ```
  29. For public repository packaging:
  30. ```text
  31. Use $content-to-skill to turn this workflow memo into an open-source skill repository with README, LICENSE, validation script, and retest prompts.
  32. ```
  33. ## What It Produces
  34. Minimum skill package:
  35. ```txt
  36. skill-name/
  37. ├── SKILL.md
  38. ├── agents/openai.yaml
  39. └── examples/retest-prompts.md
  40. ```
  41. For richer sources it may also add:
  42. ```txt
  43. references/
  44. scripts/
  45. test-prompts.json
  46. README.md
  47. LICENSE
  48. .gitignore
  49. ```
  50. ## Safety Rule
  51. Content To Skill must name the evidence boundary before turning content into instructions. If a video, document, repository, or private source is unavailable, the generated skill must say so and either ask for accessible material or mark the result as dry-run.
  52. ## Validate
  53. Run the local package validator:
  54. ```bash
  55. scripts/check_skill_package.py .
  56. ```
  57. If this repository is nested inside a larger skill repository that has a generic quality checker, also run that checker against this folder.
  58. ## License
  59. MIT. See [LICENSE](LICENSE).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.