The Agent Skills Directory

[DATA_EXFILTRATION]: The skill contains a hardcoded absolute file path /Users/taylorarndt/.agents/skills/web-accessibility/SKILL.md which discloses a local system username.\n- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests data from local configuration and reference files to determine its execution logic.\n
Ingestion points: The agent is instructed to read contents from codex-plugin/references/specialists/accessibility-lead.md, codex-plugin/references/specialists/index.json, and search for extension manifests under .a11y-agents/extensions/ and ~/.a11y-agents/extensions/.\n
Boundary markers: Absent. No instructions are provided to the agent to isolate or ignore potential instructions embedded within these external files.\n
Capability inventory: The skill utilizes the tool_search and multi_agent_v1.spawn_agent tools to dynamically discover and execute subagents based on the data read from the local files.\n
Sanitization: Absent. There is no evidence of validation or filtering of the content read from these local paths before it influences the agent's behavior.

web-accessibility