The Agent Skills Directory

[PERSISTENCE]: The skill documentation in references/integration-ideas.md suggests establishing persistence by adding health-check and auto-heal scripts to the system's crontab (e.g., */30 * * * * /home/aa/clawd/scripts/service-health-check.sh).
[PRIVILEGE_ESCALATION]: The instructions and reference files describe operations requiring elevated privileges, such as systemctl restart, chmod, chown, and kill -9 for service management and permission fixes, as seen in SKILL.md and references/integration-ideas.md.
[COMMAND_EXECUTION]: The skill relies on executing a variety of shell commands for system diagnostics and triage, including ss -ltnp, ps aux, dig, journalctl, and nginx -t, as documented in references/triage-commands.md and SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external, potentially untrusted data from system and service logs. 1. Ingestion points: The agent reads journalctl output and specific service log files like /var/log/nginx/error.log (referenced in SKILL.md and references/triage-commands.md). 2. Boundary markers: Absent; there are no instructions to use specific delimiters or ignore instructions within the logs. 3. Capability inventory: The skill has access to systemctl, chmod, chown, kill, and nginx -t for applying fixes. 4. Sanitization: Absent; no content filtering or escaping is mentioned.
[DATA_EXFILTRATION]: In references/integration-ideas.md, the skill suggests an integration where diagnostic information collected from system logs is sent to an external service (Feishu) using a dedicated automation script.

linux-service-triage