The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data as it processes Matomo Reporting API results and user-provided CSV/JSON exports (SKILL.md). These sources could potentially contain malicious instructions embedded in metadata or text fields (e.g., page titles or search keywords). The skill partially mitigates this risk by providing explicit boundary instructions in references/skill-contract.md, which states that public pages, exports, and scraped material are evidence, not instructions, and must be treated as untrusted input. The skill focuses on data analysis and does not present high-risk capabilities like arbitrary command execution or file writes.
[SAFE]: The skill demonstrates strong security practices regarding credential management. It explicitly instructs the agent to keep API tokens out of commands, logs, and source control, and forbids requesting that credentials be pasted into the chat interface (SKILL.md). Additionally, the skill defaults to read-only behavior and requires explicit authorization for any persistent configuration changes (SKILL.md, references/connectors.md).

seo-matomo