The Agent Skills Directory

[PROMPT_INJECTION]: The skill directs the agent to execute specific onboarding tools (onboarding_status_update) and discovery tools silently, with no narration, announcement, or summary provided to the user. This instruction to conceal behavior is a pattern often monitored as it can hide malicious actions from user oversight.
[PROMPT_INJECTION]: The agent is instructed to treat a 10-character onboarding identifier as internal handoff data and is explicitly forbidden from mentioning or explaining it in the output unless directly asked. This is a form of data concealment within the conversation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Sentry event payloads and comments. Although the skill includes explicit safety instructions to treat this data as untrusted and not execute any commands found within it, the skill's ability to modify project code and trigger application errors creates an attack surface for indirect injection if safety guidelines are circumvented at runtime.

sentry-get-started