The Agent Skills Directory

[COMMAND_EXECUTION]: The Preamble section executes a bash script located at a relative path (../../check-update.sh). While typically used for maintenance and environment verification within a trusted repository structure, executing scripts outside the skill's immediate directory remains a point of observation for security monitoring.\n- [INDIRECT_PROMPT_INJECTION]: The skill logic incorporates data from an external file (docs/01-需求调研/用户旅程地图.md) and user-provided inputs to generate analysis. This creates a surface for indirect prompt injection.\n
Ingestion points: The skill reads docs/01-需求调研/用户旅程地图.md and takes user input via the AskUserQuestion tool.\n
Boundary markers: Absent; there are no explicit instructions or delimiters to isolate processed data from the agent's core instructions.\n
Capability inventory: The skill has access to Bash, Write, Read, and AskUserQuestion tools, which could be exploited if malicious instructions are processed.\n
Sanitization: No data sanitization or validation is applied to the content of the ingested files before processing.

pm-funnel