The Agent Skills Directory

[PROMPT_INJECTION]: The skill begins with a directive ("STOP -- DO NOT READ THIS FILE") that attempts to override the agent's normal processing flow by claiming the prompt was "injected" and instructing the agent to jump directly to later steps.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform file operations and run the mempalace CLI tool. It builds shell commands by interpolating variables such as <artifact-path>, <room>, and <wing> into commands like cp "<artifact-path>" "$ROOM_DIR/<basename>". This pattern presents a potential command injection surface if the agent does not strictly sanitize the values of these placeholders.
[EXTERNAL_DOWNLOADS]: The skill references several documentation URLs hosted at mempalaceofficial.com. These are informational references to a third-party site and do not involve the automated download or execution of external code.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting external data without sufficient controls.
Ingestion points: The skill reads project artifacts including CONTEXT.md, PLAN.md, and SUMMARY.md in Step 2.
Boundary markers: No specific delimiters or safety warnings are provided to prevent the agent from obeying instructions that might be embedded within these processed artifacts.
Capability inventory: The skill possesses the Bash tool for file system manipulation and CLI execution, and the Read tool for accessing local files.
Sanitization: No sanitization, escaping, or validation logic is defined for the content read from external artifacts before it is used in subsequent operations.

gsd-mempalace-capture