ldilov Recursive Structured Analysis
Trigger Signals
- the task already needs a recursive session and now needs one bounded analysis step inside that session
- the repository is mixed-language or architecture-heavy and support truth should be checked before deeper investigation
- the operator needs a durable, inspectable analysis run rather than a prompt transcript
Inspect First
.hforge/runtime/recursive/language-capabilities.json.hforge/runtime/recursive/sessions/<sessionId>/execution-policy.json.hforge/runtime/recursive/sessions/<sessionId>/capabilities.json.hforge/generated/agent-command-catalog.json
Workflow
- inspect recursive capability truth before assuming a language or adapter is available
- inspect the active recursive session and execution policy
- use one bounded structured run at a time
- inspect run history and run results instead of relying on chat-only memory
- keep support claims honest when a language is degraded, unavailable, or host-executed only
Command Surface
hforge recursive capabilities --root . --jsonhforge recursive run <sessionId> --file <snippet> --root . --jsonhforge recursive run <sessionId> --stdin --root . --jsonhforge recursive runs <sessionId> --root . --jsonhforge recursive inspect-run <sessionId> <runId> --root . --json
Operating Rule
Recursive structured analysis is the promoted execution surface for recursive repository investigation. Treat it as session-scoped, policy-bound, and artifact-backed. Do not describe it as a generic REPL or as native execution parity across all languages unless the capability map explicitly supports that claim.