The Agent Skills Directory

[SAFE]: The skill focuses on establishing secure CI/CD practices and does not contain malicious code, hidden instructions, or exfiltration patterns. It explicitly warns against embedding credentials in artifacts and emphasizes the need for explicit user authorization before performing any external mutations.
[INDIRECT_PROMPT_INJECTION]: The skill instructions define workflows for processing external delivery artifacts and pipeline configurations, which represents an inherent attack surface for indirect prompt injection via untrusted configuration data.
Ingestion points: The agent is instructed to read and modify authoritative pipeline, build, packaging, and release configuration files as noted in SKILL.md and its references.
Boundary markers: Absent in technical terms; however, the skill provides strong behavioral constraints (e.g., "Do not infer authorization for publishing... from a request to prepare or review delivery files") to mitigate accidental obedience to embedded instructions.
Capability inventory: The skill description and instructions support designing and implementing deployment workflows, which implies the agent uses tools for file system modification and potentially shell execution.
Sanitization: Absent; the skill relies on manual verification steps and explicit validation checkpoints described in the guidelines rather than automated technical sanitization.

superior-delivery