The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): No malicious instructions such as 'ignore previous instructions' or attempts to bypass safety filters were found. The skill uses standard instructional prompts to define its persona as an onboarding expert.
[DATA_EXFILTRATION] (SAFE): The skill does not perform any network operations. While it attempts to read a local file (.claude/product-marketing-context.md), this is a known pattern for contextualizing agent behavior with project-specific data and does not involve accessing sensitive system files or credentials.
[REMOTE_CODE_EXECUTION] (SAFE): No mechanisms for downloading or executing remote code were identified. There are no shell commands, curl/wget calls, or piped execution patterns.
[COMMAND_EXECUTION] (SAFE): The skill consists entirely of Markdown documentation and does not invoke any subprocesses or system-level commands.
[CREDENTIALS_UNSAFE] (SAFE): No hardcoded API keys, tokens, or secrets were found in the metadata or the reference files.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill possesses an ingestion point via the .claude/product-marketing-context.md file. However, since the skill has no dangerous capabilities (no network access, no file-writing, no command execution), the risk of an indirect prompt injection causing harm is negligible. The agent simply uses the file's content to tailor its marketing advice.

onboarding-cro