The Agent Skills Directory

[METADATA_POISONING]: The skill frontmatter lists 'Alireza Rezvani' as the author, which contradicts the system-identified author account 'ciciliaeth'. Misleading metadata is a known technique for obscuring the true origin of a skill.
[PROMPT_INJECTION]: The documentation encourages the use of high-autonomy tools and flags (e.g., codex --full-auto), which can result in the agent executing actions or commands without human review.
[INDIRECT_PROMPT_INJECTION]: The skill includes tools to analyze and score untrusted external content, such as blog posts and ad copy, which creates a vector where malicious instructions embedded in that data could influence the agent's behavior.
Ingestion points: Input files including article.md, draft.md, and copy.txt processed by the bundled Python scripts.
Boundary markers: No specific delimiters or safety warnings are provided to instruct the agent to ignore instructions embedded in the external data.
Capability inventory: The skill coordinates 27 Python scripts designed for content analysis, pricing modeling, and tracking plan generation.
Sanitization: There is no evidence of input validation or sanitization for the data processed by the tools.
[COMMAND_EXECUTION]: The documentation provides explicit shell command examples for executing Python scripts (e.g., python3 marketing-skill/content-production/scripts/content_scorer.py), instructing the user or agent to run code within the local environment.

marketing-skills