The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's preamble executes an external shell script located outside the skill's own directory.
Evidence: bash "$(dirname "${BASH_SOURCE[0]}")/../../check-update.sh" 2>/dev/null || true in SKILL.md.
Impact: While intended for update checks, this executes code whose content is not defined within the skill file itself.
[PROMPT_INJECTION]: The skill ingests data from a local file which presents an indirect prompt injection surface.
Ingestion points: docs/01-需求调研/用户旅程地图.md (read via shell check).
Boundary markers: Absent; the skill assumes the content of the document is valid data for funnel analysis.
Capability inventory: Write (to docs/05-产品战略/漏斗优化方案.md), Bash (preamble execution).
Sanitization: Absent; the skill does not explicitly sanitize the contents of the journey map before using it to drive the analysis logic.

pm-funnel