The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill directs the agent to read multiple sensitive API tokens, including PSQUARED_CRM_TOKEN, NUXT_MCP_DEMO_TOKEN, and OPENBRAND_API_KEY, from a .env file located at the repository root or current working directory. Accessing .env files is a sensitive operation as they often contain credentials for various local and remote services.
[COMMAND_EXECUTION]: The skill uses curl within the shell to interact with the vendor's APIs at crm.psquared.dev and notifications.psquared.dev. It also invokes several specialized internal agent commands, such as /setup-email-drafts, /review-demos, and /plan-campaign, to orchestrate the outreach pipeline.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from CRM opportunities and scraped website content to generate email outreach drafts, creating a surface for indirect prompt injection.
Ingestion points: Data returned from the crm.psquared.dev and notifications.psquared.dev APIs, as well as website content scraped by haiku subagents during the research phase.
Boundary markers: The instructions do not specify any delimiters or explicit instructions to treat the ingested data as untrusted content.
Capability inventory: The skill has network access via curl, file system read access for configuration files, and the ability to delegate tasks to other agent models.
Sanitization: Although the skill includes a post-generation "sanity-check" tool to verify draft health and requires human approval before sending, it lacks upfront sanitization or validation of the input data used during the generation process.

inbox-outreach-pilot