The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes a "dump" of raw external material (news, social threads, competitor moves). This untrusted data could contain malicious instructions designed to influence the agent's behavior during the triage process.
Ingestion points: The --signals argument and the "dump" described in the Inputs section of SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the input data.
Capability inventory: The skill performs text transformation and reads local brand profile files.
Sanitization: No sanitization or validation of the input content is implemented.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses brand profile data stored in ~/.claude-marketing/brands/{slug}/. Access to files in hidden home directories is a sensitive operation, though this specific path appears intended for the skill's own configuration management. No network operations were found that would exfiltrate the contents of these files.

signal-mine