The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data and interpolates it into card templates using Markdown rendering, creating a surface for indirect instructions to influence the agent or deceive users.
Ingestion points: Untrusted data is injected into templates via placeholders like {{description}} and {{impact}} in references/card-alert.json, {{anomalies_section}} and {{actions_list}} in references/card-morning-briefing.json, and {{details}} in references/card-skill-test.json.
Boundary markers: Absent. The templates do not use delimiters or instructions to ignore embedded commands within the interpolated data.
Capability inventory: The skill can execute local scripts for authentication and perform network operations (POST/PATCH) to the Feishu API.
Sanitization: Absent. There is no evidence of escaping or filtering logic for the content injected into the lark_md blocks.
[COMMAND_EXECUTION]: The skill relies on a local script, ./scripts/get_feishu_token.sh, which is executed to manage Feishu authentication tokens. This is a standard practice for this skill's use case but involves shell command invocation.

feishu-card