# Troubleshooting Start with the command's exit code and the run `manifest.json`. A console summary is useful for navigation, but the manifest records the plan, terminal statuses, artifact catalog, and actionable collector-level failure details. ## The CLI says it must run in a virtual environment The normal CLI deliberately rejects a system interpreter. Bootstrap or repair the environment with the installer, then use the explicit managed interpreter: ```powershell python -m logicytics.cli.installer # or .\.venv\Scripts\Activate.ps1 to activate if already installed python -m logicytics preflight ``` If PowerShell blocks `.\.venv\Scripts\Activate.ps1`, activation is optional. Keep using `.\.venv\Scripts\python.exe` rather than changing execution policy. ## Preflight reports invalid collectors or extensions Read the named source and diagnostic row. Then run a fresh validation after the source is corrected: ```powershell python -m logicytics preflight --invalidate-cache ``` Do not bypass preflight or manually mark a source valid. Syntax and import errors, unsafe names, missing or nondeterministic metadata, unsupported platforms, undeclared capabilities, and import-time side effects are contract violations. An optional Windows facility that is merely unavailable should be recorded distinctly from an invalid source. ## The plan is empty or a collector is not selected First confirm the exact dotted ID in [Core Collector Catalog](CORE_COLLECTORS.md) and inspect the selected mode with `--modes`. Then rebuild the plan with the same mode, profile, includes, excludes, extension switches, and capability blocks as the intended run. A collector can be excluded by a selector, not belong to the mode, be invalid in preflight, or require a capability you blocked. Plugins require `--plugins`. Their absence from a normal core-only plan is intentional. Review their source and metadata before enabling them. ## A collector is skipped, partial, failed, or canceled Read its manifest record's summary, structured error details, remediation, and retry-safe indication. Skipped commonly means a required command, optional Windows feature, device, permission, or prerequisite was unavailable. Partial means some registered artifacts may still be valuable. A failed or canceled collector does not normally erase independent collector results. To narrow a reproducible problem, run the exact collector or use sequential execution: ```powershell python -m logicytics collector core.system.system_info ` --acknowledge-authorization python -m logicytics run --mode standard --sequential ` --acknowledge-authorization ``` Do not retry a sensitive or high-impact collector until its scope is still authorized. Elevation, if required by the collector, must be granted through your normal administrative process. ## Nothing was written or packaging is absent Verify the configured `runtime.output_root`, project permissions, free disk space, and `maximum_run_output_bytes`. Relative output paths are kept inside the project and unsafe paths are rejected. A `--no-package` request intentionally omits a ZIP; the manifest-backed run directory remains the result. Use the diagnostic command for a bounded environment report: ```powershell python -m logicytics debug ``` Inspect the manifest before deleting or rerunning anything. Packaging is not a success proxy: a ZIP can coexist with partial results. ## Timeout, memory, or output limits The runtime records the termination reason. Reduce the request to the smallest needed collector, make scheduling deterministic with `--sequential`, and inspect the collector's documented bounds. Adjust collector settings only when the larger collection is authorized and the destination can safely hold the output. Do not raise limits blindly for a source that may traverse, copy, or expose sensitive data. ## Safe support material Use `debug` and a sanitized manifest excerpt. Remove credentials, key material, browser files, usernames, hostnames, paths, addresses, SIDs, and raw evidence. Never attach an entire output directory or package to a public issue. See [Evidence Review](EVIDENCE_REVIEW.md) for the handoff checklist.