The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits information to the Linear platform via the Linear MCP or gh CLI. This behavior is consistent with the skill's primary purpose of filing SPIKE issues. It incorporates a safety checkpoint by prompting the user for confirmation before filing the issue.
[CREDENTIALS_UNSAFE]: The instructions include a hardcoded email address (william@chimeranext.io) and a Linear user ID (8f14370d-3602-49e3-81f2-eeb05b965687). These are vendor-owned resources used as identifiers for task assignment rather than sensitive secrets like API keys or passwords.
[COMMAND_EXECUTION]: The skill suggests using the gh CLI with Linear integration to file issues manually. This command execution is a standard development tool and is gated by user involvement.
[PROMPT_INJECTION]: The skill ingests data from external files (.md artifacts) during the evidence gathering phase. This provides a vector for indirect prompt injection if those files contain malicious instructions. The risk is considered low because the output is reviewed by the user and targets a project management system rather than an execution environment.

feature-to-spike