The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from a vast array of external sources including social media, news sites, and research databases. This data is then summarized for the user, presenting a significant surface for indirect prompt injection where malicious instructions hidden in external content could potentially override agent behavior.
Ingestion points: External data is fetched through various opencli commands (e.g., in SKILL.md and references/finance-sources.md).
Boundary markers: Absent. Instructions focus on summarizing JSON output without specific delimiters to isolate untrusted data.
Capability inventory: Extensive shell command execution capabilities via opencli across all referenced scripts.
Sanitization: Absent. The skill summarizes data directly from the tool's JSON output without validation.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file includes an environment check performed using the !command`` syntax, which executes diagnostic shell commands (such as opencli doctor) immediately when the skill is loaded by the agent to determine setup status.
[COMMAND_EXECUTION]: The skill functions by constructing and executing shell commands through the opencli utility, using subcommands and flags discovered at runtime from the tool's registry based on user requests.
[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions for installing external software including the core @jackwener/opencli package and supplementary plugins to enable the tool's functionality.

opencli-reader