The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides instructions to access sensitive configuration files and secrets from remote production environments. In references/common-scenarios.md, commands like ssh prod 'cat /app/.env' and ssh prod 'env | sort | grep -v SECRET' are explicitly provided for comparing environments.
[PRIVILEGE_ESCALATION]: The skill instructs the agent to use sudo to gain elevated privileges for system and network diagnostic tasks. For example, references/tool-specific.md includes commands such as sudo tcpdump -i any port 80 -A and sudo dtruss -p PID.
[EXTERNAL_DOWNLOADS]: The skill suggests installing various third-party debugging and profiling tools from public registries and unknown repositories. Evidence in references/common-scenarios.md and references/tool-specific.md includes npm install -g clinic, pip install py-spy, pip install memray, and go get github.com/sasha-s/go-deadlock.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its processing of external, untrusted data combined with shell access.
Ingestion points: The skill reads data from API responses via curl, application logs via jq, and database activity via pg_stat_activity.
Boundary markers: None are defined to separate untrusted external data from agent instructions.
Capability inventory: The skill has access to the Bash tool for executing shell commands.
Sanitization: There is no evidence of sanitization or validation of external content before it is processed or used in commands.
[COMMAND_EXECUTION]: The skill relies extensively on the Bash tool for system diagnostics, including network monitoring, automated git bisect workflows, and process manipulation.

debug-ops