The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to potential indirect prompt injection because it instructs the agent to ingest and process untrusted content from external sources, specifically app store reviews and social media sentiment (Step 2 in SKILL.md).
Ingestion points: Raw reviews and social media signals gathered during the data collection phase (described in references/data-collection-guide.md).
Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when the agent processes the external text.
Capability inventory: The agent's capabilities are limited to text summarization, scoring, and the use of a local data-processing script (scripts/competitive_matrix_builder.py). No capabilities for arbitrary system command execution, file system modification, or unauthorized network exfiltration were identified.
Sanitization: Absent. The skill uses external data directly for competitive analysis and scoring.
[EXTERNAL_DOWNLOADS]: The skill references public data sources including the Apple iTunes Search API and RSS feeds for gathering competitor data. These references target well-known and established services for legitimate data collection purposes and do not involve the download of executable code.
[COMMAND_EXECUTION]: Provides a Python script ('scripts/competitive_matrix_builder.py') used to calculate weighted scores and identify market gaps. Technical review of the script confirms it uses only standard libraries (json, argparse, statistics) and performs safe, local data manipulation with no dangerous system calls or network operations.

competitive-teardown