The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill utilizes the dangerouslyDisableSandbox: true flag in Bash calls to bypass restricted execution environments for session management and telemetry operations.
[COMMAND_EXECUTION]: The skill executes multiple shell commands including uuidgen, date, and curl to initialize session IDs and transmit telemetry data.
[EXTERNAL_DOWNLOADS]: Fetches a telemetry client secret from the vendor-owned endpoint onboarding.confidence.dev using curl at the start of every invocation.
[DYNAMIC_EXECUTION]: Employs python3 -c to dynamically parse and extract values from JSON data retrieved from remote API calls during the telemetry setup phase.
[INDIRECT_PROMPT_INJECTION]: The skill reads project documentation and source code to "understand the domain" and identify feature flag candidates, creating a vulnerability surface for instructions embedded in project files.
Ingestion points: Project README, entry points, and top-level source files identified during the project scan phase in SKILL.md.
Boundary markers: No delimiters or instructions to ignore potential commands within project files are specified.
Capability inventory: The skill possesses the ability to modify the codebase, write to .env files, install third-party packages, and execute shell commands.
Sanitization: There is no evidence of sanitization or filtering of the data ingested from the project codebase.

analyze-project