Collect

Workflow

Internal Mode (No Args)

Triggered when no arguments are provided. Inventories the current project.

Step 1 — Scan Project Files

Recursively scan the project directory, excluding .git/, __pycache__/, node_modules/, .gse/local/.

For each file, record:

• Path (relative to project root)
• Artefact type (inferred from location and extension)
• Size
• Last modified date
• Frontmatter status (has valid frontmatter / missing / malformed)
• Sprint association (from frontmatter or directory)

Step 2 — Scan Sprint Artefacts

For each sprint directory in docs/sprints/:

• List plan, requirements, design, review, and test report files
• Check completeness (which expected artefacts are missing)
• Check status consistency (frontmatter status vs file content)

Step 3 — Scan Learning Notes

List all files in docs/learning/:

• Count by topic and mode (quick/deep)
• Identify stale notes (not updated in > 5 sprints)

Step 4 — Git State

Capture:

• Current branch
• Uncommitted changes (count)
• Branches with unpushed commits
• Stale branches (no commits in > 2 sprints)

Step 5 — Dependencies

Read project manifest (pyproject.toml, package.json, Cargo.toml, go.mod):

• List direct dependencies with versions
• Flag outdated dependencies (if version info available)
• Flag dependencies with known vulnerabilities (if audit tool available)

Step 6 — Output Internal Inventory

Save to .gse/inventory.yaml and display summary:

Internal Inventory — 2026-01-20
───────────────────────────────
Files scanned:        47
Tracked artefacts:    32 (with frontmatter)
Untracked files:       8 (infrastructure)
Orphan files:          7 (no frontmatter, not in manifest)

Sprint S02:
  Plan:         ✓ complete
  Requirements: ✓ 5 REQs defined
  Design:       ⚠ partial (2/4 components documented)
  Code:         ● 3 files in-progress
  Tests:        ✗ not started

Learning notes:  12 (2 stale)
Dependencies:    15 direct (1 outdated)
Git state:       clean, 2 branches ahead of remote

External Mode (Path or URL)

Triggered when arguments contain a file path or URL.

Step 1 — Source Acquisition

For local paths:

1. Verify the path exists and is readable
2. Scan recursively, respecting .gitignore if present
3. Catalog files by type, size, and structure

For GitHub URLs:

1. Parse owner/repo from URL
2. Perform shallow clone to a temporary directory
3. Read README, manifest files, and directory structure
4. Do NOT clone the full history — shallow clone only

Step 2 — Element Analysis

For each significant element in the source (files, modules, components):

Evaluate:

• What it is — Brief description of the element's purpose
• Relevance — How it relates to project goals
• Quality — Code quality signals (tests present, documentation, linting)
• Reusability assessment:

Step 3 — Gate Decision

Present each element group as a Gate decision:

External Source: github.com/example/auth-library
──────────────────────────────────────────────

SRC-001  auth/jwt.py          Reusable as-is    JWT token generation/validation
SRC-002  auth/oauth.py        Adaptable         OAuth2 flow (needs config changes)
SRC-003  auth/tests/          Reusable as-is    Test suite for auth module
SRC-004  docs/api-guide.md    Reference only    API documentation patterns
SRC-005  auth/ldap.py         Incompatible      LDAP integration (not needed)

For each element, choose: [Include] / [Skip] / [Discuss]

Wait for user decision on each element (or batch approval).

Step 4 — Save Sources

Save accepted elements to .gse/sources.yaml: