The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user input via the AskUserQuestion tool and interpolates these values directly into subagent prompts and file-writing operations without sanitization.
Ingestion points: User responses collected in Step 1 and Step 2 (variables INTERVIEW_GOAL, HYPOTHESIS, SEGMENTS, SAMPLE_PLAN) in SKILL.md.
Boundary markers: Absent. The strings are interpolated into the Task tool prompt and the Write tool content without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill utilizes Write for file system modification and Task for subagent delegation.
Sanitization: No validation or escaping is applied to the user-provided data before it is used in downstream prompts.
[COMMAND_EXECUTION]: The preamble executes a shell command to perform an update check and create a project directory.
Evidence: bash "$(dirname "${BASH_SOURCE[0]}")/../../check-update.sh" 2>/dev/null || true and mkdir -p docs/01-需求调研.
[DYNAMIC_EXECUTION]: The update script path is resolved dynamically at runtime using shell variable expansion.
Evidence: Use of $(dirname "${BASH_SOURCE[0]}") to locate check-update.sh relative to the skill's installation path.

pm-interview