The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by reading and processing terminal output from other agents.
Ingestion points: The superset terminals read command in SKILL.md fetches the content of agent terminal sessions for analysis.
Boundary markers: There are no explicit boundary markers or instructions to treat the ingested terminal content as untrusted data or to ignore potential instructions embedded within it.
Capability inventory: The skill uses the superset CLI to list and read workspaces, tasks, and terminals. While the instructions state the digest is read-only, it suggests following up with terminal inputs or cleanup actions as separate steps, which could be influenced by injected content.
Sanitization: No formal sanitization or filtering is applied to the terminal output before the agent processes it to classify states or extract relevant lines.

superset-standup