The Agent Skills Directory

[SAFE]: The skill uses dynamic context injection to retrieve the current Git branch name for project grounding. This is a read-only, benign system command used according to standard development practices.
[SAFE]: The skill ingests user-provided research topics through arguments and interpolates them into instructions for subagents. The potential for indirect prompt injection is mitigated by the skill's design, which emphasizes execution isolation and mandatory primary-source verification to prevent ungrounded synthesis.
[SAFE]: All tool invocations (Agent, Workflow) and file system interactions follow the defined research discipline and do not involve unauthorized credential access, external downloads from untrusted sources, or persistent system changes.

research-deep