メニュー
Normalize SKILL.md artifacts into Scheduling-Structural-Logical (SSL) JSON representations using a conservative multi-pass extraction pipeline.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/github/gh-aw --skill ssl-skill-normalizerコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
GitHub Agentic Workflows (`gh-aw`) is a GitHub CLI extension for writing Agentic Workflows in markdown and compiling them to GitHub Actions.
Operate safely and efficiently inside a gh-aw workflow with a restricted tools/bash allowlist, and correctly triage tool-denial events before they exhaust the session's denial budget.
Optional Sergo examples for cache formats and reporting templates.
SKILL.md を表示中
SOC 職業分類に基づく
| name | ssl-skill-normalizer |
| description | Normalize SKILL.md artifacts into Scheduling-Structural-Logical (SSL) JSON representations using a conservative multi-pass extraction pipeline. |
| tools | ["read_file","write_file","search_files","json_validate","create_artifact","run_tests"] |
| inputs | ["skill_path"] |
| outputs | ["ssl_json","validation_report"] |
This skill converts markdown-based skill artifacts into a structured Scheduling-Structural-Logical (SSL) representation as introduced in:
Liang et al., "From Skill Text to Skill Structure: The Scheduling-Structural-Logical Representation for Agent Skills", arXiv:2604.24026 (2026).
SSL addresses the core limitation of free-form skill text: it is human-readable but hard for agents to reason over, discover, and audit. By mapping each skill into three complementary layers, SSL makes skills searchable (improved MRR 0.573 → 0.707 in the paper) and risk-assessable (improved macro F1 0.744 → 0.787).
The representation is grounded in Schank & Abelson's theories of Memory Organization Packets (MOPs), Script Theory, and Conceptual Dependency. Each layer captures a different dimension of skill knowledge:
Answers: When should this skill be invoked? By whom, given which inputs and outputs?
Fields extracted:
id — stable lowercase identifiername — human-readable skill namegoal — one-sentence purposeintent_signature — typed function signature (fn($input) -> $output)inputs — $-prefixed named input bindingsoutputs — $-prefixed named output bindingsdependencies — explicit runtime tool or library requirementscontrol_flow_features — e.g. sequential, conditional, loopentry_scene — ID of the first scene to executesubscene_refs — IDs of any nested/delegated scenesAnswers: What are the macro-level execution stages and how do they connect?
Each scene is a named execution stage with:
id — unique within the skilltype — one of the restricted scene-type enum (see below)goal — what the scene accomplishesentry_condition — precondition for entering the sceneexit_condition — postcondition that must hold on exitnext_scene_rules — conditional transitions to the next scene ID, END_SUCCESS, or END_FAILinputs / outputs — $-prefixed bindings consumed and producedentry_logic_step — ID of the first logic step in this sceneAnswers: What atomic operations are performed, on which resources?
Each logic step is an indivisible operation with:
id — unique within the skillscene_id — owning sceneaction_type — one of the restricted action-type enum (see below)resource_scope — one of the restricted resource-scope enum (see below)description — one sentence describing the operationinputs / outputs — named $-variable bindingsnext — ID of the following step, YIELD_SUCCESS, or YIELD_FAIL| Value | Meaning |
|---|---|
PREPARE | Setup: load inputs, configure environment |
ACQUIRE | Receive or fetch required data |
REASON | Analyze, infer, or plan |
ACT | Produce or transform primary output |
VERIFY | Validate outputs or preconditions |
RECOVER | Handle failure; retry or compensate |
FINALIZE | Write results, emit notifications, clean up |
| Value | Meaning |
|---|---|
READ | Consume data from a resource without side effects |
SELECT | Choose among alternatives |
COMPARE | Diff or rank two or more values |
VALIDATE | Assert a constraint or schema |
INFER | Derive new information via reasoning |
WRITE | Produce or overwrite data in a resource |
UPDATE_STATE | Mutate shared state |
CALL_TOOL | Invoke an external tool or subprocess |
REQUEST | Send a request to an external service |
TRANSFER | Move data between resources |
NOTIFY | Emit a message or event |
TERMINATE | End execution and return control |
| Value | Meaning |
|---|---|
MEMORY | In-process working memory |
LOCAL_FS | Local file system |
CODEBASE | Source code under version control |
PROCESS | OS process or shell |
USER_DATA | User-provided or personal data |
CREDENTIALS | Secrets, tokens, or credentials |
NETWORK | Remote network resource |
OTHER | Any resource not covered above |
END_SUCCESS | END_FAILYIELD_SUCCESS | YIELD_FAILnull, empty arrays, or coarse-grained classifications when evidence is weak.Read the source SKILL.md, then extract the scheduling layer.
Produce scheduling with all fields in Layer 1. When evidence is absent for an optional field, emit an empty array or null.
Requirements
snake_case.Analyse the skill's execution flow and decompose it into macro-level scenes.
Requirements
Constraints
next_scene_rules target must resolve to another scene ID, END_SUCCESS, or END_FAIL.RECOVER scene when the source describes retry or error-recovery behaviour.Expand each scene into its sequence of atomic logic steps.
Split a step whenever any of the following changes:
Requirements
$-prefixed variable bindings for all named data ($user_request, $selected_file, $generated_output).Validate the draft SSL JSON against all of the following rules:
| Rule | Check |
|---|---|
| JSON syntax | Well-formed JSON |
| Required fields | All top-level fields present |
| Enum membership | All enum fields use allowed values only |
| Unique identifiers | All scene IDs and step IDs are globally unique |
| Entry pointer | entry_scene references an existing scene ID |
| Scene entry pointer | entry_logic_step references an existing step ID |
| Scene containment | All referenced scene IDs exist |
| Logic-step containment | All referenced step IDs exist |
| Transition validity | All transition targets are valid scene/step IDs or terminal values |
| Graph integrity | No unreachable scenes or dangling references |
Failure Handling
Generate a normalization report containing:
Include per-artifact diagnostics with the specific Pass-4 rule that caused rejection.
Do not expose secrets or credentials in reports.
The skill succeeds when:
The skill fails when:
A schema-valid SSL JSON file named ssl.json placed alongside the source SKILL.md. Top-level keys: scheduling, scenes, logic_steps.
A validation and normalization report summarizing accepted artifacts, rejected artifacts, per-artifact validation diagnostics, and retry behaviour.
To apply this skill to a SKILL.md artifact:
skill_path pointing to the target SKILL.md.RECOVER pass retries generation up to the retry budget.ssl.json is written alongside the source file.validation_report output to confirm acceptance.For batch normalization, invoke this skill once per artifact and aggregate the per-artifact reports.