The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands to process project data and manage the system clipboard.
Evidence includes use of git log, git shortlog, awk, and shell piping logic for data transformation.
Specifically uses ~/bin/mdcopy or pandoc combined with xclip to transfer data to the clipboard.
[DATA_EXFILTRATION]: The skill is designed to read and aggregate data from production environments (SQL databases and Git history) and transfer it to the system clipboard.
It accesses potentially sensitive information such as user email addresses, event logs, and error strings from production tables (e.g., usage_events, users, job_runs).
While the skill includes instructions to prioritize secure secret resolution (e.g., via fnox) and read-only access, the data it aggregates is intended to be moved out of the agent's restricted context to the user's clipboard.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes data from external, untrusted sources.
Ingestion points: Reads commit messages from git log and error strings/logs from the job_runs table in production.
Boundary markers: No explicit boundary markers or isolation instructions are provided for the interpolated data in the report generation phase.
Capability inventory: The agent has capabilities to execute shell commands and perform database queries.
Sanitization: Error strings are truncated to 90 characters, but no content sanitization or escaping is performed on commit messages or error descriptions before they are interpreted or included in the final report.

activity-report