The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes shell commands via the execFileSync module, including git, gh, and headless Chrome for PDF generation. It also heavily relies on the external screenless CLI tool to manage background waiting, mailing, and placing calls.
[PRIVILEGE_ESCALATION]: The SKILL.md instructions explicitly use dangerouslyDisableSandbox: true when initializing the background waiter process (screenless wait). This configuration bypasses standard execution restrictions to allow the background task to persist and perform operations outside the typical agent sandbox.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources which could be manipulated to influence the agent's behavior during report generation or brief preparation.
Ingestion points: Data is collected from pull request titles, bodies, and diffs via gh, as well as ticket titles and descriptions from external trackers like Linear or Jira.
Boundary markers: None. The skill does not instruct the agent to use delimiters or specific safety warnings to ignore embedded instructions within the ingested data.
Capability inventory: The skill has capabilities to execute shell commands, write files to local directories, and trigger network-based notifications (sending emails and placing calls) through the vendor-provided screenless tool.
Sanitization: None. The collected data is processed and incorporated directly into call briefs and JSON manifests without explicit filtering or escaping.
[DATA_EXPOSURE]: The skill performs extensive reading of private repository content, pull request history, and project management metadata to generate its 'deep dive' and status reports.
[PERSISTENCE_MECHANISMS]: The skill utilizes the agent session's /loop heartbeat feature combined with a backgrounded screenless wait process to maintain execution across time, ensuring the nightly analysis occurs even if the user is not actively interacting with the agent.

morning-pr-review