القائمة
Use when resolving language-specific skills for a workflow. Triggers on multi-language projects or when language-specific conventions are needed before implementation.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Use when producing product artifacts - epics, user stories, acceptance criteria, or analytics specs.
Use when defining product direction, vision document sections, or roadmap. Triggers on strategic product planning, vision creation, or vision document quality review.
Use when designing UI, choosing colors, typography, spacing, creating visual hierarchy, or making design decisions. Based on Refactoring UI by Wathan & Schoger.
Use when someone needs a plain-language explanation of how an existing feature or edge case behaves in the codebase. Triggers - non-technical stakeholder asks how something works, need code-informed functional explanation, need to understand rules or edge cases without implementation planning.
Use when a developer wants to go from rough idea to working code autonomously. Triggers - imperfect prompt, solo dev wants full pipeline, "just build it", idea to implementation without manual steps.
Use when creating a new agentic workflow skill. Triggers - need new multi-step orchestrated process, want to add workflow to agentic framework, building spec-driven pipeline.
| name | agentic:skill:skill-injection-protocol |
| description | Use when resolving language-specific skills for a workflow. Triggers on multi-language projects or when language-specific conventions are needed before implementation. |
Resolve language-specific skills based on project tech stack and profile configuration. The orchestrator executes this protocol ONCE per workflow and caches the result as language_skills_prompt in workflow-state.yaml.
A language_skills_prompt string value:
""Read .agentic.settings.json from the IDE directory. Extract the profiles array.
{
"profiles": [
{ "name": "typescript", "detect": ["typescript", "ts", "node", "bun", "deno"], "skills": ["typescript-engineer"] },
{ "name": "python", "detect": ["python", "py", "pip", "uv", "poetry", "conda"], "skills": ["python-engineer"] }
]
}
If profiles is missing or empty → language_skills_prompt: "", stop.
Read technical-context.md from the workflow output path. Extract tech_stack from YAML frontmatter.
---
tech_stack: [typescript, react, node]
---
If technical-context.md doesn't exist or tech_stack is missing → language_skills_prompt: "", stop.
For each tech_stack entry, search all profiles' detect arrays (case-insensitive). Collect matched profiles' skills. Deduplicate.
Example: tech_stack [typescript, react, node] → all match typescript profile → skills: ["typescript-engineer"]
Check skillOverrides in settings:
| Override | Example | Effect |
|---|---|---|
| Replace | {"typescript-engineer": "custom-ts"} | Load custom-ts instead |
| Remove | {"typescript-engineer": "_remove_"} | Don't load that skill |
If tech_stack entries have no matching profile:
decision-log.md as LOW_CONFIDENCEIf no skills matched → language_skills_prompt: "", stop.
Set language_skills_prompt to:
---
# Language Skills
Load these language-specific skills before proceeding:
- Skill(skill="{skill-1}")
- Skill(skill="{skill-2}")
Confirm: "Language skills loaded: {skill-1}, {skill-2}"
---
Cache in workflow-state.yaml:
language_skills_prompt: |
---
# Language Skills
Load these language-specific skills before proceeding:
- Skill(skill="{skill-1}")
- Skill(skill="{skill-2}")
Confirm: "Language skills loaded: {skill-1}, {skill-2}"
---
| Mistake | Fix |
|---|---|
| Agent tries to execute this protocol | Only orchestrator executes this. Agents receive {language_skills_prompt} in their Task() prompt. |
| Resolving per-step instead of once | Resolve once at the workflow resolution point, cache in workflow-state.yaml. |
| Wrong skill name prefix | Use short names: typescript-engineer, not agentic-skill-typescript-engineer |