The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the repository being audited, which presents a surface for indirect prompt injection.
Ingestion points: The skill uses the Read tool to ingest the content of all .go files within the target repository for analysis (references/phase-2-dispatch-agents.md).
Boundary markers: There are no explicit instructions or delimiters provided to the agent to differentiate between the code being analyzed and potential instructions embedded within comments or strings in that code.
Capability inventory: The skill possesses a broad set of capabilities, including Bash, Write, Edit, Task, and Skill tools, which could be targeted by an injection attack (SKILL.md).
Sanitization: The skill does not implement sanitization or validation of the code content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill utilizes shell commands via the Bash tool for repository discovery and dependency mapping.
The discovery process involves running commands like find, grep, and sed to identify package structures and file counts (references/phase-1-discover-commands.md).
Sub-agents are instructed to use grep to identify function usage and dead code across the repository (references/phase-2-dispatch-agents.md).
[DYNAMIC_EXECUTION]: The skill uses the Task tool to dynamically dispatch sub-agents (golang-general-engineer) to perform parallel reviews of segmented package groups (SKILL.md, references/phase-2-dispatch-agents.md).

sapcc-audit