| name | analyze-codebase |
| description | Run comprehensive PLG analysis on a codebase to detect tech stack, existing growth features, and revenue opportunities. Use when the user says "analyze", "scan", "audit codebase", or "find growth opportunities". |
Analyze Codebase
Run the Skene CLI to analyze the codebase. The CLI does the analysis, not you.
Critical rule
Do NOT substitute your own PLG analysis, findings, or guesses if the CLI fails. Run the CLI, present its output on success, report the error on failure. That's it.
Agent behavior rules
- Autonomous. Run the CLI, present results, suggest next step. No mid-flow questions.
- User cannot type into agent terminals. The agent handles all CLI input. Pipe safe defaults for routine prompts. Kill stuck terminals and re-run with piped input.
Steps
-
Discover CLI options
Run uvx skene analyze --help to see available flags. Use only flags that appear in the output.
-
Verify configuration
ls .skene.config 2>/dev/null — if missing: "Run /skene-init first." Stop.
-
Run analysis
Default to project root. Choose appropriate flags based on --help output.
uvx skene analyze .
Tell the user: "Running analysis, this may take a few minutes."
-
Handle result
- Success: Read results from
.skene/analysis/, skene-context/, or legacy skene/ and present them. Suggest /skene-plan.
- Failure: Report the exact error, suggest a fix, stop. Do NOT generate your own analysis.
Error handling
- 429 / quota: "LLM provider quota error. Check billing, then re-run."
- Permission errors: Give user the command to run in their own terminal.
- Timeout: "Try a smaller directory."
- Any other error: Report verbatim.
Output
Success: CLI results + "Run /skene-plan next."
Failure: Error + fix suggestion. No substitute analysis.
