The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains instructions to execute shell commands that incorporate user-supplied variables including <user-domain>, <top-competitor>, and <niche>.
Evidence: python scripts/dfs_client.py competitors --target "<user-domain>" and python scripts/dfs_client.py ideas --keyword "<niche>".
Risk: The pattern of wrapping user input in double quotes within a shell command is susceptible to command injection. If a user provides a value like example.com" ; malicious_command #, the agent would execute the malicious command in the host environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external script outputs and incorporates it into the agent's reasoning process without sanitization.
Ingestion points: The output of scripts/dfs_client.py (specifically keyword ideas and competitor ranked data) is used to build the 'Entity map' and 'Subtopic inventory' in SKILL.md.
Boundary markers: Absent. The instructions do not define delimiters for external data or advise the agent to disregard instructions embedded within the SEO data.
Capability inventory: The skill has the capability to write files to the local filesystem and execute a follow-up script (scripts/report_publish.py) that handles the generated content.
Sanitization: Absent. No filtering, validation, or escaping is performed on the data retrieved from the external SEO analysis before it is processed by the LLM.

topical-authority-planner