القائمة
Call-chain tracing skill template. Guides creation of project-specific call-chain skills for tracing API data flow, verifying type matching, and checking endpoint correctness.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
AI Agent Skills creation methodology. Triggered when creating skill systems for projects, generating project-specific skills, understanding skill templates, model tier delegation, or validating skill accuracy. META skill — creates other skills as output. Deploy: copy this directory to .claude/skills/skill-builder/.
Code map skill template. Used to guide creation of project-specific code map skills. Triggered when asking about file locations, directory structure, component navigation, or route mapping. L0 — Must be executed by Haiku, pure file path lookup.
Delegation rules skill template. Used to guide creation of project-specific delegation rules. Defines which tasks must be delegated to Haiku, delegation format, and sub-agent output standards.
Development Standards skill template. Guides creation of project-specific development standards skills. Triggers on questions about tech stack, API conventions, code standards, coding style, or layered architecture.
Development workflow skill template. Guides creation of project-specific workflow skills. Triggered when asking about development steps, phase order, or checklists for common tasks.
Little Notebook — records user dissatisfaction and satisfaction moments during agent interaction. Use when user expresses anger, frustration, abusive language, or strong satisfaction. L0 — delegate to Haiku for recording and cleanup.
استنادا إلى تصنيف SOC المهني
| name | example-call-chain |
| description | Call-chain tracing skill template. Guides creation of project-specific call-chain skills for tracing API data flow, verifying type matching, and checking endpoint correctness. |
| model_tier | L1 |
| skill_tier | functional |
| version | 1.0.0 |
| status | active |
牵一发而动全身。 ——龚自珍《自春徂秋偶有所触》 改一行代码之前先知道谁在调用它。这不是顾虑，是常识。
Positioning: Functional tier skill — provides structured call-chain tracing methodology for verifying data flow correctness. Orchestrated by delegation (planning) and skill-builder-guide (meta).
Step 1: Identify entry point
├─ API endpoint / event handler / scheduled task / message consumer
└─ Record: input types, validation rules
Step 2: Trace through layers
├─ Entry layer → Business layer → Data layer → External calls
└─ Record at each layer: method name, parameter types, return type
Step 3: Verify final call
├─ SQL query / external API call / message publish / cache write
└─ Check: parameters correctly assembled, types match
Step 4: Error path check
├─ Exceptions caught at each layer?
└─ Error response includes correct status code?
## {Endpoint Name} Call Chain
{Caller}
│ {HTTP method} {path}
▼
{Controller}.{method}({param}: {type})
│ → calls {Service}.{method}({param}: {type})
▼
{Service}.{method}({param}: {type})
│ → business logic: {summary}
│ → calls {Repository}.{method}({query})
▼
{Repository}.{method}({query})
│ → {SQL / external API / message}
▼
{Database / External Service}
| # | Layer Transition | Check | Status |
|---|---|---|---|
| 1 | Request → Controller | Param types match DTO | ✅/❌ |
| 2 | Controller → Service | Service param = DTO or derived type | ✅/❌ |
| 3 | Service → Repository | Query params match Prisma/SQL types | ✅/❌ |
| 4 | Service → External | API contract types match | ✅/❌ |
| 5 | Service → Response | Return type matches API schema | ✅/❌ |
When generating a project-specific call-chain skill, map the project's actual layer names to the tracing template. Include project-specific external API call signatures, error handling patterns, and one worked example of a real endpoint's full chain.
Architecture-specific tracing patterns (CLI, event-driven, frontend, layered backend): references/tracing-patterns.md.
After completing this skill, recommend the next skill based on output characteristics:
| Condition | Recommend |
|---|---|
| New change requirement found | {project}-change-model |
L1 — Sonnet / functional tier: Requires code-level tracing and type-compatibility reasoning. File lookups delegated to L0 — Haiku.