The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to ingest untrusted data from a user-provided specification document (as seen in Step 4 of SKILL.md). This content is analyzed to populate a PROJECT.md file, which then acts as the primary context for an architect sub-agent to generate a project roadmap. Malicious instructions hidden within the specification document could potentially influence the agent's behavior during the analysis or planning phases.
Ingestion points: The spec file is read and parsed in Step 4 of SKILL.md.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the imported content.
Capability inventory: The skill can create directories (mkdir), copy files (cp), perform git operations (git add/commit), and update internal state via state-write.sh.
Sanitization: No sanitization of the specification content is described before it is interpolated into project artifacts.
[COMMAND_EXECUTION]: The skill executes several shell commands to manage project artifacts, including mkdir, cp, git add, and git commit (Steps 5 and 7). It uses variables such as <spec-file-path> and <spec-filename> which are derived from user input or file system search results. While these operations are standard for project management tooling, they represent a potential attack surface if the execution environment does not properly escape or handle these variables.

import-spec-file