The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the longbridge CLI to fetch financial records and perform community interactions. These actions are consistent with the skill's documented purpose of stock analysis.
[PROMPT_INJECTION]: The skill metadata contains a discrepancy, labeling the risk level as read_only despite the presence of content-creation commands in the reference files (longbridge topic create, longbridge topic create-reply). This misleading classification could lead to a misjudgment of the agent's capabilities regarding active interactions in community forums.
[PROMPT_INJECTION]: The skill handles untrusted data from news and community sources, creating an indirect prompt injection surface.
Ingestion points: External data is ingested into the agent context via longbridge news and longbridge topic commands.
Boundary markers: The instructions lack specific delimiters or instructions to isolate external data from the system prompt.
Capability inventory: The skill includes functional tools for publishing content to external forums, which could be triggered by instructions embedded in the ingested data.
Sanitization: Basic filtering is applied to specific speculative keywords, but there is no general mechanism to sanitize or escape malicious instructions in the external content.

longbridge-content