The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to enumerate contents of the local file system to identify other installed skills.
Evidence: The 'Companion check' section in SKILL.md directs the agent to run ls ~/.claude/skills/ ~/.agent/skills/ ~/.agents/skills/ .claude/skills/ .agent/skills/ .agents/skills/ 2>/dev/null.
[PROMPT_INJECTION]: Contains automated instructions that trigger specific behaviors and system commands immediately upon skill activation.
Evidence: The 'Companion check' logic in SKILL.md forces the agent to perform a filesystem scan and offer to install additional software without a direct user request for that action.
[EXTERNAL_DOWNLOADS]: Recommends the installation of external code from a third-party source using a package runner.
Evidence: Both README.md and SKILL.md suggest using npx skills add AbsolutelySkilled/AbsolutelySkilled to download and install the primary skill and its recommended companions.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted user-provided data without specified sanitization.
Ingestion points: Processes survey responses, interview transcripts, and behavioral analytics data provided by the user (referencing SKILL.md).
Boundary markers: Absent; there are no instructions to ignore embedded commands in the data being processed.
Capability inventory: Access to the shell via ls and tool invocation capabilities.
Sanitization: Absent; no logic is provided to escape or validate external content before processing.

customer-research