Agnostic Prompt Standard (APS) v1.0 Skill
The reference framework to generate, compile, and lint greenfield prompts that conform to the Agnostic Prompt Standard (APS) v1.0.
Purely documentation-based skill: no script execution, no external side effects, no sensitive-data handling demands, MIT licensed. However, metadata attributes authorship to Christopher Buckley and points to chris-buckley/agnostic-prompt-standard while the host repo is jsburckhardt's fork; the provenance chain is unverified, and the publisher is unverified. The normative references/ documents were not provided in evidence, so the actual content of 06-logging-and-privacy cannot be confirmed. Deducted for unverifiable attribution and missing core normative material.
SKILL.md is a structurally self-consistent entry document; assets (constant blocks, format contracts) are syntactically uniform and consistently named. But claimed lint/compile capability depends on scripts/ (declared empty by default), the core references/ documents are absent from evidence, and the repository CI tests cover devcontainer features, not this skill. Deducted for missing key-path material and no executable failure-feedback path.
Target scenario is clear (generate, compile, lint APS v1.0-conformant prompts); the entry doc distinguishes normative vs non-normative material and provides multi-platform adapters and a scaffold template. However, trigger conditions, non-fit boundaries, and Chinese-language support are undeclared, and usability depends on references/ content not supplied. Deducted for limited evidence on boundaries and trigger precision.
Good layered information architecture (entry → spec → assets → processes → guides → platform adapters), version metadata (spec_version, framework_revision), template skeleton, and license declared. But last_updated is a future date (2026-03-12), authorship metadata conflicts with the host repository, there is no CHANGELOG or maintenance-responsibility statement, and scripts/ is an empty placeholder. Deducted for metadata credibility flaws and incomplete versioning/governance.
As a prompt-standard framework, its value lies in consistent prompt structure and format contracts; the rich assets of format and constant examples have direct reference value. But it cannot be statically verified that prompts generated under this standard outperform hand-written ones, the normative core (references/) is absent, and output usability is unconfirmed. Deducted for unverifiable core value claim and unproven comparative benefit.
Only internal consistency is checkable (asset naming, version numbers, layout claims match). The spec documents are missing, there are no tests, no third-party execution evidence, and the README badge pointing to external releases cannot be corroborated. Deducted because key claims (spec content, lint capability) are independently unverifiable and evidence types are thin.
- Authorship metadata (Christopher Buckley / chris-buckley repo) conflicts with the host repository jsburckhardt/devcontainer-features, and last_updated is a future date (2026-03-12); verify the true origin before use.
- The normative core references/ documents and lint scripts are absent from evidence or empty placeholders; as shipped, the skill is effectively an entry index and its conformance-checking capability is unproven.
- Repository CI only tests devcontainer features, not this skill; there are no APS-specific tests or third-party execution evidence.
- No declared Chinese-language support or trigger boundaries; semantic invocation may misfire.
What does this skill do, and when should you use it?
This skill, located at .github/skills/agnostic-prompt-standard/ in the jsburckhardt/devcontainer-features repository, is the entrypoint for the Agnostic Prompt Standard (APS) v1.0. It ships the eight normative spec documents in references/, covering structure, vocabulary, linting and formatting, agentic control, schemas and types, grammar, logging and privacy, and error taxonomy. Around the spec it provides reusable examples (format and constants blocks), a skill scaffolding template, non-normative platform adapters (VS Code + GitHub Copilot, Claude Code, OpenCode, and generic/MCP), plus guides for skill building, authoring, and subagent architecture. It suits anyone who wants to write portable, verifiable prompts against a formal standard. Note: while the spec describes what an LSP/linter should ingest, the bundled scripts/ directory is empty by default — no working lint/compile tool is included.
As a skill entrypoint it performs no computation; it provides: 1) eight normative APS v1.0 documents under references/ defining prompt structure, vocabulary, linting and formatting, agentic control, schemas and types, grammar, logging and privacy, and an error taxonomy; 2) reusable assets including three constants-block examples (JSON, text, CSV) and ten format-block examples (code changes, code map, docs index, error, hierarchical outline, ideation list, link manifest, Markdown table, SMEAC plan, API coverage table), plus two composite examples; 3) a minimal skill skeleton in _template/ for scaffolding new skills; 4) the build-skill.md process document and three guides (skill authoring, MCP tool bridge, cross-platform subagent architecture); 5) non-normative platform adapters under platforms/, each with platform constants, tool registry, and format contracts.
- A prompt engineer writing greenfield prompts who wants consistent structure, vocabulary, and error handling enforced by a formal spec.
- A skill author targeting multiple platforms (VS Code Copilot, Claude Code, OpenCode, MCP tool layers) who needs ready-made adapter contracts.
- A team that wants to scaffold new skills with a uniform structure using the _template/ skeleton.
- A developer building an LSP or linter for prompts who needs a spec corpus to ingest from references/.
- An agent developer looking for portable coordinator/worker subagent architecture guidance.
What are this skill's strengths and limitations?
- Complete eight-part normative spec covering structure, vocabulary, linting, agentic control, schemas, grammar, privacy, and error taxonomy.
- Ten format-block and three constants-block examples reduce onboarding effort.
- Four platform adapters (VS Code Copilot, Claude Code, OpenCode, generic/MCP) support cross-platform authoring.
- Includes a scaffolding template and a subagent architecture guide for standardized team output.
- MIT licensed, free to use and modify.
- scripts/ is empty by default: the advertised 'compile and lint' capability ships with no actual tooling.
- The repository README describes an unrelated devcontainer features collection, making the documentation context confusing.
- SKILL.md metadata points to chris-buckley/agnostic-prompt-standard, a different repository from where the skill is hosted; provenance needs verification.
- The last_updated date reads 2026-03-12, a future date, so metadata reliability should be questioned.
- The spec corpus is heavyweight; lightweight use cases may find it over-engineered.
How do you install this skill?
Copy the .github/skills/agnostic-prompt-standard/ folder into your Agent Skills-compatible client's skills directory (e.g. Claude Code's skills folder). SKILL.md is the entrypoint. No package-manager or one-line install command is documented in the source.
How do you use this skill?
After reading SKILL.md, consult as needed: references/ contains the spec documents defining how to write APS-compliant prompts; assets/ provides directly reusable format and constants block examples; processes/build-skill.md describes the workflow for building new APS-compliant skills; use _template/ to scaffold new skills; see platforms/<platform>/adaptor.md for platform-specific constants, tool registries, and format contracts. The scripts/ directory is empty by default — there are no runnable compile/lint commands.