Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/ChatGPT-SKILL-SYSTEM/llms.txt

Use this file to discover all available pages before exploring further.

skill-adapter is a built-in skill that helps you inspect an external skill ZIP before adapting it to the GPT Project Skill System package format. Rather than silently converting or guessing at compatibility, it produces a structured candidate package plan and a compatibility report with a clear rating — PASS, PASS_WITH_WARNINGS, or BLOCKED — so you know exactly what needs operator attention before you build a new package. It ships as dist/skills/skill-adapter.zip and is ready to add to any Project alongside dist/GPT.SKILLS.zip.

​
Loading skill-adapter

Add dist/skills/skill-adapter.zip to your ChatGPT Project sources, then run the two standard lifecycle commands at the start of your session:
1

Unpack skill-adapter

Installs the skill files into the runtime environment.
SKILL skill-adapter UNPACK
2

Load skill-adapter

Mounts the skill and makes it semantically active for the current session.
SKILL skill-adapter LOAD
Once loaded, you can give skill-adapter a path to an external skill ZIP (or an extracted folder) and ask it to inspect the package.

​
What skill-adapter checks

During a mechanical inspection, skill-adapter examines the following aspects of a candidate skill package:

Path safety

Flags absolute paths, .. traversal, backslashes, drive letters, empty segments, and single wrapper directories at the ZIP root.

ZIP entries

Verifies that all ZIP entries use POSIX / separators. Backslash entries are a hard fail.

File inventory

Walks the full extracted file tree and categorises every file as a likely support, tool, or asset file.

Candidate manifests

Reads an existing MANIFEST.json if present and validates it against the system rules. Reports whether it can be reused, needs correction, or must be replaced.

Likely primary file

Searches for SKILL.md, skill.md, README.md, or the first Markdown file to propose a primary_file candidate.

Scripts

Identifies script files (.py, .sh, .js, etc.) and — for Python — parses imports to flag non-standard-library dependencies.

Binary files

Detects non-textual assets and notes any that are referenced in load_sequence, which is a hard blocker.

External mechanism indicators

Scans text files for markers such as pip install, npm install, daemon, watcher, webhook, and source-runtime wording (e.g. claude) that indicate incompatible runtime assumptions.

​
Compatibility ratings

Every report produced by skill-adapter ends with one of three ratings:

PASS

No hard blockers found. The package can be built under the current rules. Any remaining assumptions are stated explicitly in the report.

PASS_WITH_WARNINGS

The package can be built, but some files or behaviors require operator review or task-time caution before the skill is used in production.

BLOCKED

A hard blocker prevents a valid package candidate. The report lists every blocker and, where possible, describes the fix required.

​
Hard blockers

The following conditions always produce a BLOCKED rating. None of them can be overridden by operator decision alone — the underlying issue must be fixed in the source package first.
  • Unsafe paths — absolute paths, .. traversal, backslashes, drive letters, or empty path segments anywhere in the package.
  • ZIP entries with backslashes — POSIX / is required for all ZIP entry names.
  • Missing declared files — a file referenced in load_sequence, support_files, tool_files, or asset_files does not exist after extraction.
  • Invalid manifest identity — skill_name does not match the expected slug, or required manifest fields are missing or the wrong type.
  • Mounted file not readable as UTF-8 — any file in load_sequence that cannot be decoded as UTF-8 text.
  • Script or binary file in load_sequence — only AI-only UTF-8 textual files may be semantically mounted.
  • Required dependency unavailable in the GPT Python environment — tool files that depend on packages not available in the GPT Python environment.
  • Required hook, daemon, watcher, external CLI, local app, or network install — the skill depends on runtime mechanisms the system does not provide.

​
Report output shape

Every adaptation report follows this structure: 
Compatibility: PASS | PASS_WITH_WARNINGS | BLOCKED

Source:
- path:
- shape:

Loader mode:
- fallback | explicit

Manifest draft:
  {
    "skill_name": "...",
    "version": "...",
    "primary_file": "...",
    "load_sequence": ["..."],
    "support_files": ["..."],
    "tool_files": ["..."],
    "asset_files": []
  }

Mounted files:
- <relative path to each file in load_sequence>

Physical files:
- support_files: <list>
- tool_files:    <list>
- asset_files:   <list>

Findings:
- severity: BLOCKER | WARNING
  file: <path>
  issue: <description>
  fix: <recommended action>

Operator decisions:
- <items that require a human call before the package can be finalised>

Claim:
- candidate inspected for GPT Project Skill System adaptation;
  operator review still required
Loader mode is either fallback (root SKILL.md with minimal manifest) or explicit (declared primary_file and load_sequence). The Manifest draft section is omitted when blockers prevent a valid draft from being produced.

​
skill-adapter manifest

The full MANIFEST.json shipped with the skill-adapter package: 
{
  "skill_name": "skill-adapter",
  "version": "0.1.0",
  "primary_file": "SKILL.md",
  "load_sequence": ["SKILL.md"],
  "support_files": ["references/adaptation_report_schema.md"],
  "tool_files": ["scripts/inspect_skill_candidate.py"],
  "asset_files": [],
  "capabilities": ["skill-adaptation", "compatibility-review", "manifest-drafting"],
  "runtime_hints": {
    "uses_python": true,
    "output": "candidate package plan plus compatibility report"
  }
}
SKILL.md is the only semantically mounted file. references/adaptation_report_schema.md is a support reference consulted when drafting reports. scripts/inspect_skill_candidate.py is a tool file — physically available after unpack, but never run automatically.
When Code Interpreter file access is available in your session, you can ask skill-adapter to run scripts/inspect_skill_candidate.py directly against a candidate ZIP or extracted folder. The script produces a full JSON inspection report covering path safety, wrapper detection, manifest validation, Python import analysis, and runtime-marker scanning — giving you a precise, machine-checked baseline before any manual adaptation work begins.
skill-adapter operates entirely within the bounds of the current session. It does not modify the system core, does not make any inspected or adapted package active, does not execute tool files automatically, and does not install any dependencies. A compatibility report is evidence for operator review — it is not a guarantee of general compatibility or a substitute for testing the adapted package yourself.

Build docs developers (and LLMs) love

Get started for freeTalk to us