The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands for path discovery and project analysis. This includes an aggressive file system search using find $HOME to locate its internal scripts and the execution of project-specific CLI tools like kubectl, pulumi, terraform, and flux when the optional 'live mode' is enabled for infrastructure projects.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted documentation files from the target repository, such as README.md, CLAUDE.md, and ADRs. Malicious instructions embedded in these files could attempt to manipulate the agent's reasoning or the content of the generated reports, despite instructions to prioritize source code over documentation.
[DATA_EXFILTRATION]: When operating in 'live mode', the skill accesses potentially sensitive infrastructure configurations and metadata. Although the instructions explicitly forbid revealing secret values and mandate read-only access, the inherent risk of processing live environment data and writing it to local documentation files could lead to accidental exposure of system topology or metadata if the agent's redaction logic fails.

system-flow-mapper