The Agent Skills Directory

[PROMPT_INJECTION]: The instructions in SKILL.md are strictly professional and focus on survey methodology. There are no attempts to override safety filters, bypass instructions, or activate hidden 'developer' modes.
[DATA_EXFILTRATION]: No patterns of hardcoded credentials, sensitive file path access, or network transmission calls (such as curl or wget) were found. The skill operates entirely within the text-processing context of the agent.
[COMMAND_EXECUTION]: The skill does not contain any shell commands, script execution logic, or system calls. The dynamic context injection pattern (!command) is not used.
[REMOTE_CODE_EXECUTION]: The skill only references internal markdown files (references/question-types.md and references/bias-checklist.md). It does not download or execute any remote scripts or packages.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from users, such as survey goals and pre-existing questionnaires.
Ingestion points: SKILL.md identifies user descriptions of survey goals and uploaded files as primary inputs.
Boundary markers: There are no explicit delimiters or boundary markers defined in the instructions for handling these inputs.
Capability inventory: The skill is limited to generating text-based survey designs and audit reports; it lacks capabilities for file writing, network access, or command execution.
Sanitization: No explicit sanitization or filtering of user input is mentioned. However, since the output is purely informational text and no dangerous tools are available to the skill, the surface for exploitation is negligible.

pm-survey-designer