The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from previous session logs and the active conversation context to create distilled summaries. This creates an attack surface where malicious instructions embedded in those sources could potentially influence the agent during the distillation process.
Ingestion points: Reads files from the quality_reports/session_logs/ directory and accesses the current conversation history in Step 1 and Step 2.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the ingested text as untrusted data or to ignore embedded directives.
Capability inventory: The skill has access to powerful tools including Read, Write, Glob, Grep, and Bash to perform its tasks.
Sanitization: The instructions do not define any sanitization, filtering, or validation steps for the content retrieved from session logs.
[COMMAND_EXECUTION]: The skill instructions direct the agent to use the Bash tool to execute shell commands (such as git commands for branch name and last commit SHA) to gather environmental metadata for the session report. While these are legitimate functional requirements for the skill, they involve direct shell execution.

compress-session