The Agent Skills Directory

[SAFE]: The skill was audited for prompt injection patterns. The instructions use natural language to define a persona and workflow; no attempts to bypass safety filters, override system prompts, or extract instructions were detected.
[SAFE]: Data exposure and exfiltration check: The skill does not access sensitive file paths (e.g., .ssh, .aws) and contains no network operations (curl, wget, etc.).
[SAFE]: Obfuscation check: No Base64, hex encoding, zero-width characters, or homoglyph substitutions were found in the instructions or example data.
[SAFE]: Unverifiable dependencies and remote code execution check: The skill does not include any package installation commands or remote script execution patterns. It relies entirely on internal logic and sub-skills.
[SAFE]: Dynamic context injection check: The skill does not use the dynamic context execution syntax (!command) to run shell commands at load time.
[SAFE]: Indirect prompt injection check: While the skill ingests external data (task lists, PRDs), it lacks exploitable capabilities such as file writing, network access, or code execution. Furthermore, it explicitly mandates human review gates ('HARD-GATE') before proceeding between phases, effectively mitigating risks associated with untrusted data ingestion.

project-manager