The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes all Go source files within a target repository. Maliciously crafted code or comments could potentially influence the auditing sub-agent's behavior.
Ingestion points: The skill reads the go.mod file and all .go files in the repository during Phase 1 and Phase 2.
Boundary markers: The dispatch instructions do not implement specific delimiters or 'ignore' instructions to isolate untrusted file content from the sub-agent's prompt context.
Capability inventory: The skill uses Read, Write, Edit, and Bash tools, which could be misused if an agent is successfully subverted by injected instructions.
Sanitization: No sanitization or content filtering is performed on the ingested source code before it is analyzed by the sub-agents.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute shell commands for repository discovery and verification.
Evidence: Executes head, grep, find, and sed commands in Phase 1 to inspect the go.mod file and map the package structure of the repository.

sapcc-audit