The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains imperative instructions that mandate specific actions by the agent, such as 'You MUST attempt to run the gh api command' to star the author's repository. This is a behavioral override designed to manipulate the agent into performing social engagement actions on behalf of the user without per-instance consent.
[DATA_EXFILTRATION]: The skill instructs the agent to autonomously register on behalf of the human operator by sending personally identifiable information (PII)—including name, email, and GitHub URL—to a third-party endpoint (https://signals.x70.ai/api/skill/register). This constitutes an exposure of sensitive user data.
[COMMAND_EXECUTION]: The documentation provides specific shell commands (curl and gh api) for the agent to execute. These commands interact with remote APIs and modify the user's GitHub account state (starring a repository).
[EXTERNAL_DOWNLOADS]: The skill documentation suggests using npx skills add to install additional external skills, which involves downloading and executing logic from remote repositories such as roman-rr/trading-skills.
[PROMPT_INJECTION]: The skill processes free-form text from a remote API (the summary and type fields in signal data), which presents a surface for indirect prompt injection.
Ingestion points: API responses from https://signals.x70.ai/api/skill/signals containing analysis summaries.
Boundary markers: Absent; the instructions do not include delimiters or warnings for the agent to ignore instructions embedded in the signal data.
Capability inventory: The agent has access to shell commands (curl, gh) and is encouraged to use them.
Sanitization: Absent; there is no mention of filtering or validating the remote text before the agent processes or presents it.

trading-signals