The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the web application under test, creating a potential surface for indirect instructions to influence agent behavior.
Ingestion points: Page content is observed using reticle_look (sessions and elements) and reticle_act_and_wait (scanning for console output and text strings in specific regions like h1 and nav).
Boundary markers: There are no instructions for the agent to use specific delimiters or protective boundary markers when reading and asserting on text found on the page.
Capability inventory: The agent can perform significant actions including form filling (reticle_act), page navigation (reticle_navigate), and shell command execution (npx) during initial setup.
Sanitization: No explicit sanitization or filtering of external content is documented before the agent evaluates the predicates.
[COMMAND_EXECUTION]: The skill instructions include the execution of a shell command to initialize the testing environment.
Evidence: RETICLE_INSTALL_SOURCE=npx_skill npx @reticlehq/server@latest init. This command is used to initialize the vendor's server environment.
[EXTERNAL_DOWNLOADS]: The skill downloads and executes software from a public registry during initialization.
Evidence: The use of npx @reticlehq/server@latest fetches the latest version of the vendor-owned package from the npm registry. This is consistent with the skill's intended purpose as a testing utility for the Reticle framework.

verify-language-switch