The Agent Skills Directory

[SAFE]: The skill does not contain any malicious code, prompt injections, or obfuscated instructions. It focuses on recovery workflows with multiple layers of safety validation.
[COMMAND_EXECUTION]: The skill interacts with the local environment via the jianying CLI. Security is maintained through strict pre-execution checks, including version, doctor, and capabilities audits to ensure the runtime environment matches expectations before any action is taken.
[DATA_EXFILTRATION]: The instructions explicitly forbid the collection or recording of credentials. All data processing is confined to local paths and pre-defined provider boundaries, with a specific rule against原地覆盖 (overwriting original source files), ensuring data integrity and preventing unauthorized data movement.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests external project files and task logs, it mitigates risk by enforcing a structured schema (jianying-job/v2) and requiring a human-in-the-loop for ambiguous or failed states. It uses project verify to ensure structural integrity rather than relying on potentially poisoned content.

jianying-recover