oo-find-skills
Find, compare, and install published OOMOL/oo skills. Use when the user asks to find, search for, discover, recommend, compare, choose, or install an existing skill for a task; asks whether there is a skill that can do something; or explicitly mentions the OOMOL/oo skill catalog. Do not use for creating or editing local skills, generic skill design, or non-OOMOL skill catalogs.
oomol-lab · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add oomol-lab/oo-cli --skill oo-find-skills技能说明
name: oo-find-skills description: Find, compare, and install published OOMOL/oo skills. Use when the user asks to find, search for, discover, recommend, compare, choose, or install an existing skill for a task; asks whether there is a skill that can do something; or explicitly mentions the OOMOL/oo skill catalog. Do not use for creating or editing local skills, generic skill design, or non-OOMOL skill catalogs. disable-model-invocation:
allowed-tools: [Bash(oo *)]
oo Find Skills
Use this skill when the user wants to discover, compare, recommend, or install
existing published skills from the OOMOL or oo skill catalog through
oo skills search and oo skills install.
Read references/oo-cli-contract.md when you
need exact oo skills search or oo skills install command forms, JSON
expectations, output-shape rules, or failure-handling details.
Workflow
1. Normalize the request into a search sentence and keywords
Convert the user request into:
- one short internal intent statement
- one concise English sentence for the search text
1to3required keywords or short phrases for the--keywordsfilter
Rules:
- The sentence must always be in English, regardless of the user's language.
- Keywords may use the user's original language. Keep product names, brand
names, and proper nouns exactly as the user wrote them and do not translate
them. For example, keep
滴答清单as滴答清单; do not turn it intoTickTick. - The backend tokenizes
--keywords, so original-language and product-name keywords reach the catalog entry the user actually wants. The free-text sentence alone runs an untokenized semantic search that can map a localized product onto a different global product. - The sentence should describe only the user's need itself.
- Prefer a short sentence built from task + capability + domain or constraint.
- Do not add meta words such as
skill,skills,search, orinstallunless the user's actual need depends on those words.
Examples:
- Sentence:
translate scanned images from Japanese to EnglishKeywords:Japanese,image translation - Sentence:
create a task in the dida365 to-do appKeywords:滴答清单,dida365 - Sentence:
generate a QR code from textKeywords:QR code - Sentence:
convert speech to textKeywords:transcription - Sentence:
write Markdown more effectivelyKeywords:Markdown,writing
2. Search for candidate skills
Always pass the keywords through --keywords:
oo skills search "<english sentence>" --keywords "<comma-separated keywords>" --json
Then:
- Rank only from the returned JSON array.
- Prefer semantic matches between the original user goal and the returned
description,skillDisplayName, orname. - Keep one primary skill.
- Keep one fallback skill only if it is genuinely plausible.
If the returned array is empty, tell the user there is no matching skill available right now and stop. Do not present a menu.
3. Filter installable results
- Treat a search item as installable only when it includes both
packageNameandname. - If the JSON array is non-empty but no installable items exist, stop and tell the user that no installable skill could be derived from the search results. Do not present a menu.
4. Rank the installable results
- Rank the installable JSON items using only the fields in the response.
- Pick one primary skill and, only if credible, one fallback skill.
5. Report or ask the user to choose
- If the user asked to find, compare, recommend, or check whether a skill exists, report the primary skill and any credible fallback, briefly explain why they match, and ask whether the user wants to install one. Do not install yet.
- If the user explicitly asked to install, add, set up, or choose and install a matching skill, ask the user to choose between the available install actions.
- If the user later names or selects one of the reported installable skills, treat that as confirmation for that exact install action.
Interaction rules:
- If a credible fallback exists, offer these actions:
- Install the primary skill as
primarySkillName (primaryPackageName) - Install the fallback skill as
fallbackSkillName (fallbackPackageName) - Install both
- Install neither
- Install the primary skill as
- If no credible fallback exists, offer only these actions:
- Install the primary skill as
primarySkillName (primaryPackageName) - Install neither
- Install the primary skill as
- First, try to use the
<!-- agentic:var skillSelectionPromptTool -->tool with one short multiple-choice question that includes only the actions that are actually available. - If the
<!-- agentic:var skillSelectionPromptTool -->tool is unavailable in the current mode or the tool call fails, fall back to plain text. - If the
<!-- agentic:var skillSelectionPromptTool -->UI returnsNone of the above, treat that as the same outcome asInstall neither.
- Prefer asking the user with a short multiple-choice prompt in chat.
- If the host provides a short-question UI, you may use it. Otherwise ask in plain text.
- If the UI returns
None of the above, treat that as the same outcome asInstall neither.
- In either UI or text form, the label for every install action must include
the concrete
skillName (packageName)text.
Fallback text format:
1. Install <skillName> (<packageName>)
2. Install <skillName2> (<packageName2>)
3. Install both
4. Install neither
Reply with: 1, 2, 3, or 4
If no credible fallback exists, use:
1. Install <skillName> (<packageName>)
2. Install neither
Reply with: 1 or 2
If the user reply is not one of the explicit numbers that correspond to the currently available option
阅读完整来源 ↗