The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill provides a recipe for a 'Named Coprocess' in SKILL.md that executes arbitrary Python code from a pipe using the exec() function. This pattern is inherently dangerous as it allows for arbitrary code execution if the input pipe receives data from an untrusted or attacker-controlled source.
[COMMAND_EXECUTION]: The skill demonstrates patterns for sourcing the output of external commands directly into the current shell environment (e.g., source <(aws configure export-credentials --format env)). This can lead to unexpected environment modifications or the execution of malicious commands if the external tool's output is compromised.
[DATA_EXFILTRATION]: Multiple examples in SKILL.md illustrate sending data, including system log contents, to external recipients using the mail command (e.g., tee >(grep -i "CRITICAL" | mail -s "Critical Alert" admin@example.com)).
[PRIVILEGE_ESCALATION]: The skill includes examples that access sensitive system files and logs, such as /var/log/syslog and /etc/config, which typically require elevated privileges to read. An agent following these patterns might attempt to escalate its own privileges or perform unauthorized data access.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, making it vulnerable to indirect prompt injection if those sources contain malicious instructions.
Ingestion points: Data is ingested via curl from external APIs, ssh from remote servers, and tail/cat from system log files.
Boundary markers: No explicit boundary markers or 'ignore' instructions are used when interpolating external data into shell commands or processing pipelines.
Capability inventory: The skill utilizes powerful tools including bash, ssh, python3 (with dynamic execution), aws-cli, and mail.
Sanitization: The provided patterns rely on standard utilities like grep and jq for filtering but do not implement robust sanitization or validation of the ingested content.

process-substitution-fifos