The Agent Skills Directory

Data Redaction and Privacy: The skill incorporates high-quality security practices by requiring the agent to redact credentials, tokens, private endpoints, and PII from reports before they are shared. It also mandates the use of private, untracked directories for raw evidence to prevent accidental exposure.
Controlled System Interaction: The skill utilizes command-line tools and shell execution to capture system state and reproduce defects. These actions are governed by specific instructions emphasizing isolation, positive controls, and verification of fixes.
Indirect Prompt Injection Surface: Because the skill processes external content like logs and source code, it possesses a surface for indirect prompt injection. It mitigates this risk by requiring that every factual claim traces to captured evidence and by mandating the redaction of sensitive data from the final output.
Use of Standard Tooling: The skill employs standard developer tools, such as the GitHub CLI, to verify repository permissions and filing destinations, ensuring that automated actions align with user intent and repository configurations.

bug-report