[ERROR] scan failed: Anthropic provider requested but no credentials available. Pass --api-key / --oauth-token, set $ANTHROPIC_API_KEY or $CLAUDE_CODE_OAUTH_TOKEN, or run `agentgg init --provider anthropic`. | The active provider has no credential, and this run passed none. | Run agentgg init, set the environment variable, or pass --api-key or --oauth-token for this run. See Providers. |
[ERROR] scan failed: git is not installed or not on PATH. --diff requires git to enumerate changed files. | The scan ran --diff but git is not on the system PATH. | Install git, then run the scan again. See Scan a pull request. |
[ERROR] scan failed: git diff-tree <commit> failed. Check that '<commit>' is a valid commit SHA reachable in this repo. | The commit does not exist in this repository. | Confirm the ref exists with git log <commit>. See Scan a pull request. |
Done. 0 finding(s) across 0 file(s). | The root path, --exclude, --only, or --diff matched no files. | Check the root path. Loosen --exclude, --only, or the --diff range. Run agentgg status ./out to confirm the count. |
sql-injection: capped to 300 candidate file(s) (--max-files-per-agent; 40 dropped) | The agent’s candidate files went over the --max-files-per-agent cap. | Raise --max-files-per-agent, or narrow the scan with --only or --exclude. See Limit what gets scanned. |
Capping to 250 batch(es) across all agents (--max-batches; 30 dropped → 4 agent(s) capped) | The scan queued more batches than --max-batches allows. | Raise --max-batches. A dropped batch runs on the next scan against the same --output directory. |
[WARN] precondition:<slug>: check failed, running the agent without it | The model call of the agent’s precondition check failed. The agent runs anyway, so the scan loses no coverage. | Read the [ERROR] line above it for the cause. To check the precondition again, run the scan with --re-recon. |
[ERROR] scan failed: --live-validate requires --target-url | The scan ran --live-validate with no target. | Add --target-url with the root URL of your running application. See Test findings against a running app. |
Docker is not installed, so the live-validation sandbox cannot start. Install Docker Desktop or the Docker engine, then retry. | Live validation needs Docker. The scan continues and keeps the validator’s verdicts. | Install Docker, then run agentgg live-validate on the output directory. |
The Docker daemon is not responding. Start Docker and retry. | Docker is installed but does not run. The scan continues and keeps the validator’s verdicts. | Start Docker, then run agentgg live-validate on the output directory. |
Sandbox port 8931 is already in use. Stop whatever holds it (a leftover sandbox: docker ps, then docker rm -f <id>) and retry. | An earlier sandbox, or another program, uses the port of the sandbox. | Stop that container or program, then run agentgg live-validate on the output directory. |
live validation: target http://localhost:3000 did not respond, skipping | The application at --target-url did not answer for one minute. Each failed try logs live validation: target probe failed (<reason>) before this line. | Start the application, check the URL, then run agentgg live-validate on the output directory. |
[WARN] [fix:<id>] no fix: the answer did not match the repository. <problem> | The model’s edits did not match the source file two times. The finding has no suggested fix. | Run agentgg fix on the output directory to try again, or use a stronger model. See Suggested fixes. |
[WARN] Suggested fixes stopped early: <message> | A provider error, such as a rejected key or no credit, stopped the fix phase. The report is still written. | Correct the provider problem, then run agentgg fix on the output directory. |
