The Agent Skills Directory

[PROMPT_INJECTION]: The skill's instructions are focused on logical orchestration and do not attempt to manipulate the agent's safety protocols. While the skill processes external data (like logs and code), which is an inherent surface for indirect injection, the methodology's emphasis on falsification and cross-checking reduces the impact of untrusted inputs. Data ingestion points include the user observation argument and external artifacts (logs, code, configs). Boundary markers and sanitization logic are not explicitly detailed in the markdown instructions. Capability inventory is limited to tracing-specific tools like trace_timeline.- [DATA_EXFILTRATION]: There are no instructions or patterns that would lead to the unauthorized transmission of data to external domains or the extraction of sensitive credentials.- [REMOTE_CODE_EXECUTION]: No external dependencies or remote scripts are downloaded or executed by the skill.- [COMMAND_EXECUTION]: The skill describes a conceptual workflow for tracing and does not invoke arbitrary shell commands.- [NO_CODE]: The skill is comprised of instructional markdown and metadata without any accompanying scripts or binary files.

trace