The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it is designed to ingest and act upon data from external, untrusted sources within the Paperclip platform.
Ingestion points: As outlined in SKILL.md and references/api-reference.md, the skill retrieves potentially attacker-controlled data through API calls like GET /api/issues/{issueId}/comments and GET /api/issues/{issueId}/heartbeat-context.
Boundary markers: There is no evidence of boundary markers or instructions to the agent to disregard embedded directives within the processed data.
Capability inventory: The agent has broad permissions to mutate state via the Paperclip API and can perform local operations such as git worktree management and executing the paperclipai CLI tool.
Sanitization: The instructions lack any requirement for sanitizing or validating external data before it is processed or used in decision-making.
[EXTERNAL_DOWNLOADS]: The skill facilitates the import and installation of additional agent skills from external sources, including the skills.sh registry and GitHub repositories, as described in references/company-skills.md.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the paperclipai CLI tool and git (e.g., git worktree add) to manage local development environments and perform self-tests.

paperclip