Stack Detection

Use this skill to identify the repository technology stack using configuration and manifest files.

This skill does not analyze source code. It only reads configuration and manifest files.

Purpose

Produce a structured stack profile describing:

• language
• framework
• runtime
• bundler
• package manager
• tooling signals
• deployment indicators
• run commands

Gating Policy

• Cost Class: CHEAP (Low-cost).
• Mandatory Sequence: Runs only after fingerprint has completed.
• Skip Conditions:
  • Skip if material_change = false.
  • Skip if docs/STACK_PROFILE.md already exists and the repository fingerprint is unchanged.
  • Never run without fingerprint evidence.

Hard Rules

• Maximum files read: 15.
• Read policy: Only read configuration or manifest files.
• No source: Never read source directories such as src/, lib/, or components/.
• Evidence-based: Never infer technologies without direct file evidence.
• Traceability: Every detection must cite its specific source file.
• Termination: Stop reading files once 15 files have been processed.

File Priority Order

The agent must inspect files in this specific order and stop after 15 files:

1. package.json
2. pnpm-workspace.yaml
3. tsconfig.json
4. astro.config.*
5. vite.config.*
6. next.config.*
7. nuxt.config.*
8. angular.json
9. svelte.config.*
10. turbo.json
11. nx.json
12. Dockerfile
13. docker-compose.yml
14. .github/workflows/*.yml (first match)
15. netlify.toml or vercel.json
16. pyproject.toml or requirements.txt
17. Cargo.toml
18. go.mod

Detection Targets

Extract signals and provide evidence for:

• Language: (e.g., TypeScript, JavaScript, Python, Rust, Go).
• Framework: (e.g., React, Astro, Vue, Angular, Svelte).
• Runtime: (e.g., Node.js, Bun, Deno, Python runtime).
• Bundler: (e.g., Vite, Webpack, Esbuild, Turbo).
• Styling tools: (e.g., Tailwind, Sass, PostCSS).
• Package manager: (e.g., npm, yarn, pnpm, bun, pip, cargo, go modules).
• Testing framework: (e.g., Jest, Vitest, Playwright, Cypress).
• Deployment platform: (e.g., Vercel, Netlify, AWS, Docker).
• Build tooling: (e.g., ESLint, Prettier, Husky).

Run Command Extraction

Detect run commands specifically for dev, build, test, and lint.

• Primary Source: package.json scripts.
• Fallback Sources: Makefile or CI workflow commands.

Confidence & Mode Rules

Confidence Levels

• HIGH → 3 or more independent evidence signals.
• MEDIUM → 1–2 evidence signals.
• LOW → Weak or conflicting signals.

If conflicting signals appear: Mark detection as CONFLICTING, set confidence to LOW, and list both sources.

Detection Mode

Record as one of: config-evidence, partial-detection, or conflicting.

Output & Communication

The stack-analyzer is the owner of this artifact. Upon completion:

1. Persistence: Write the structured results to docs/STACK_PROFILE.md.
2. Decision Log: Append a short completion entry to docs/DECISIONS.md.
3. Communicate: Post the stack summary to the Shared Task List so the Team Lead can update project context.

Required Output Structure (docs/STACK_PROFILE.md)

• Summary: High-level overview of the detected stack.
• Confidence: HIGH | MEDIUM | LOW.
• Detection Mode: config-evidence | partial-detection | conflicting.
• Core Stack Table: | Category | Technology | Evidence File |
• Run Commands: | Purpose | Command | Source |
• Tooling and Deployment: | Category | Signal Found | Evidence File |
• Conflicts or Ambiguities: List any conflicting signals found.