akshay-na/project-identity

project-identity (Gemini)

Parity: Same resolution as Cursor ~/.gemini/skills/kb-identity/SKILL.md (after DotMate stow) → durable path ~/ai-brain/projects/<project_name>/. Keep algorithm aligned if you edit either skill.

Inputs

• project_root: absolute path to the working copy (git repo root, worktree checkout, or non-git folder).

Outputs (conceptual)

• project_name: lowercase slug (from origin URL last segment or directory basename).
• content_brain / kb_path: ~/ai-brain/projects/<project_name>/ with ~ expanded to $HOME.

Algorithm (summary)

1) Locate git config

• If project_root/.git is a directory → read project_root/.git/config.
• If project_root/.git is a file (worktree) → parse gitdir: line → find main repo’s .git/config (path up to and including .git before /worktrees/) and read that config.

2) remote_url

• Parse INI: [remote "origin"] → url = ...
• If missing → derived_from = folder_name

3) project_name

• If remote_url set → last path segment of repo path, strip .git, lowercase (HTTPS, git@host:, ssh:// — same rules as kb-identity).
• Else → basename(project_root).toLowerCase()

4) KB path

content_brain = $HOME/ai-brain/projects/<project_name>/

Use project_name only (not full_identity hash) for the directory name — matches kb-identity Step 5 / kb_path.

5) Errors

• No .git at project_root: still allow project_name = basename(project_root) for non-git folders (optional: treat as not_git_repo if policy forbids — default allow folder fallback).

Overrides (env, already applied before this skill)

Handled in rules/brain-paths.md / rules/project-identity.md:

• GEMINI_CONTENT_BRAIN — absolute path bypasses derivation.
• GEMINI_PROJECT_ROOT — forces project_root for derivation.

Caching

Agents may cache (project_root → content_brain) for the invocation; invalidate if project_root or origin changes.