The Agent Skills Directory

[COMMAND_EXECUTION]: The shell script block incorporates the user-controlled argument args.bloq directly into an environment variable assignment (BLOQ="${{args.bloq}}") and a subsequent CLI command (iris bloqs ingest "$BLOQ" "$MEM"). Although the variable is enclosed in double quotes, the pattern remains susceptible to command injection if the execution environment allows for shell expansion (such as backticks or subshells) within the substituted argument value.
[PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by persisting agent-generated outputs into a local memory file that is subsequently re-ingested for future context (RAG).
Ingestion points: Data from the entry variable is appended to the local mem file in SKILL.md and later uploaded via iris bloqs ingest.
Boundary markers: The skill lacks explicit boundary markers or instructional delimiters that would prompt the agent to ignore malicious instructions embedded within the stored memory content.
Capability inventory: The skill possesses the ability to write to the file system and perform network-based ingestion through the iris platform tools.
Sanitization: The implementation does not perform validation, filtering, or escaping of the content before it is stored and reused in the agent's knowledge base.

agentic-loop