The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from web pages and accessibility trees, which could contain malicious instructions aimed at the agent.\n
Ingestion points: The skill reads DOM content and accessibility refs via orca snapshot and orca eval in references/browser-verification.md.\n
Boundary markers: Absent. The instructions do not include delimiters or warnings to treat the browser data as untrusted.\n
Capability inventory: The skill has the capability to execute orca CLI commands, perform git operations, run python3 scripts, and write files to the .scratchpad/ directory.\n
Sanitization: Absent. There is no evidence of filtering or sanitizing the ingested web content before the agent processes it.\n- [COMMAND_EXECUTION]: The skill relies on shell command execution via the orca CLI for terminal and browser management, git for repository manipulation, and python3 for local utility tasks.\n- [DYNAMIC_EXECUTION]: The skill uses orca eval to execute JavaScript within the browser context and python3 -c to execute a script for decoding and saving screenshot data. These are used for functional automation within the Orca IDE.

orca