The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the steam-web-pp-cli and steam-web-pp-mcp tools from the author's NPM package (@mvanhorn/printing-press-library) and GitHub repository (github.com/mvanhorn/printing-press-library) using npx and go install.
[COMMAND_EXECUTION]: The skill relies on executing a local binary (steam-web-pp-cli) via the shell to perform its functions, as defined in the allowed-tools frontmatter and the execution instructions.
[DATA_EXFILTRATION]: The CLI tool supports a --deliver webhook:<url> flag, allowing it to POST query results directly to an external URL. Furthermore, it includes a feedback command that can transmit local data to a remote endpoint if the STEAM_WEB_FEEDBACK_ENDPOINT environment variable is set.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources, specifically the Steam Web API (e.g., news feeds, user reviews, and achievement metadata), which could contain adversarial content meant to influence the agent.
Ingestion points: External data from Steam APIs (including news search, reviews, and app schemas) is ingested into the agent context via the CLI commands described in SKILL.md.
Boundary markers: While the CLI wraps output in a JSON envelope (meta and results keys), there are no documented instructions for the agent to ignore potentially malicious instructions embedded within that data (absent in SKILL.md).
Capability inventory: The agent has access to Bash and Read tools, enabling it to execute the CLI and process the resulting output (SKILL.md).
Sanitization: There is no mention of sanitization, filtering, or escaping of external API content before it is delivered to the agent context (SKILL.md).

pp-steam-web