The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses local file paths containing potentially sensitive brand profiles, guidelines, and configuration data.
Evidence: Accesses ~/.claude-marketing/brands/_active-brand.json, ~/.claude-marketing/brands/{slug}/profile.json, and ~/.claude-marketing/brands/{slug}/guidelines/_manifest.json to load brand context and compliance rules.
[COMMAND_EXECUTION]: The skill references and utilizes local Python scripts for technical analysis and scraping that are not included in the provided file content, making their full behavior unverifiable.
Evidence: The skill lists tech-seo-auditor.py, content-scorer.py, and competitor-scraper.py in the 'Scripts Used' section.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as it ingests untrusted data from multiple sources to perform its audit.
Ingestion points: Processes user-provided URLs, data sources (CSV/JSON files), and API endpoints to evaluate row count, uniqueness, and content quality.
Boundary markers: There are no explicit instructions to the agent to ignore or delimit instructions that might be embedded within the audited content or data sources.
Capability inventory: The agent has the capability to run the specified analysis scripts (tech-seo-auditor.py, etc.) based on instructions derived from the processed data.
Sanitization: No sanitization, escaping, or validation of external content is specified before the data is used for assessment and template design.
[DYNAMIC_EXECUTION]: The skill relies on the execution of Python scripts at runtime to perform content scoring and technical auditing.
Evidence: Mentions content-scorer.py and tech-seo-auditor.py as primary processing tools for the auditing workflow.

programmatic-seo