The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill adds and fetches data from an external repository (defaulting to https://github.com/qwibitai/pepper.git) using git remote add and git fetch.
[REMOTE_CODE_EXECUTION]: After merging code from the remote source, the skill automatically executes npm run build and npm test. This pattern results in the execution of potentially unvetted code on the local system.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to manage the Git lifecycle (git merge, git status, git commit) and project build tools (npm).
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through its ingestion of external data.
Ingestion points: Untrusted code and instructions enter the context via git merge from an upstream repository in SKILL.md.
Boundary markers: There are no protective delimiters or instructions to prevent the agent from obeying malicious prompts hidden within the merged code or conflict markers.
Capability inventory: The skill possesses the ability to execute shell commands (git, npm) and perform file system operations, which could be abused if malicious instructions are processed.
Sanitization: No validation or sanitization of the remote content is performed before the agent attempts to resolve conflicts or run build scripts.

update-skills