| name | asa-skill-identity-decomposer |
| description | Use when you need to identify purpose, type, target agent, primary outputs, and core identity before workflow analysis. |
| internal_meta_skill | true |
| asa_role | structure_analyst |
| output_contract | structure_analysis.identity |
ASA Skill Identity Decomposer
Use this internal meta-skill when a model must determine the identity, purpose, type, target agent, and primary outputs of a skill package.
This skill fixes the failure mode where an analyst copies frontmatter or jumps directly into workflow without first explaining what the skill actually is.
Inputs
- Deterministic inventory for one skill package.
SKILL.md frontmatter and core body blocks.
- Package file list and detected resource roles.
- Existing evidence objects when available.
Process
- Read the skill name and description from frontmatter.
- Identify the user problem the skill claims to solve.
- Classify the primary skill type:
file, tool, workflow, meta, domain, governance, or unknown.
- Identify secondary tags only when supported by wording or file structure.
- Identify target agents only when explicitly named or strongly implied by source context.
- Identify primary outputs and distinguish final outputs from intermediate artifacts.
- Write one concise anatomy sentence in Chinese and English.
- Attach evidence to each high-confidence identity claim.
Output Contract
Return identity content suitable for structure_analysis.json:
identity:
one_line:
zh:
en:
skill_type:
primary:
secondary: []
target_agents: []
primary_outputs: []
value_proposition:
zh:
en:
confidence:
overall:
notes:
evidence: []
Anchor-Aware Output
In addition to identity content, emit minimal anchor cards when the source supports them. Keep this lightweight; do not generate a full composition plan from this skill.
Recommended anchor types:
identity_anchor: what the skill is and what problem it solves.
value_anchor: the primary capability or value unit worth borrowing.
output_anchor: final or intermediate output shapes.
target_user_anchor: explicit or strongly implied target agent/user.
Minimal anchor card shape:
anchors:
- id: anchor.<skill_id>.identity
source_skill_id:
anchor_type: identity_anchor
name:
summary:
evidence:
path:
quote:
confidence: low | medium | high
reuse_modes:
- reference_only
- temporary_composition
- learning_note
risks: []
Anchor rules:
- Emit only anchors supported by frontmatter, body text, or deterministic inventory.
- Use
medium or low confidence for inferred target users.
- Do not emit workflow anchors from this skill; workflow anchors belong to the workflow trace method.
- Do not plan composition here; pass anchors to
asa-anchor-composition-planner.
Evidence Rules
- Use
explicit evidence for frontmatter name, description, and directly stated outputs.
- Use
structural evidence for file-layout conclusions.
- Use
inferred evidence for target agent or value proposition when not directly stated.
- Do not mark target agents as high confidence unless source text or repository context supports it.
Failure Modes
- Do not treat every skill as a workflow.
- Do not replace identity analysis with a raw frontmatter copy.
- Do not invent audience or target agent from project popularity.
- Do not claim final outputs from examples unless the skill asks the agent to produce them.
Quality Rubric
A strong identity decomposition answers:
- What is this skill?
- Who or what is it for?
- What problem does it solve?
- What does it produce?
- What source evidence supports those claims?