Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →get-bb/secrets
Securely request API keys, access tokens, passwords, webhook secrets, or other credentials from the user and write them to a dotenv file without exposing their values to the agent.
Passed 11 of 17 test cases with the skill, and 6 of 17 without it.
and 5 more in the full report
7 fixed8 passed either way8 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 7 test cases the skill fixed. Same model, same prompt, with and without the skill.
Generate the exact command to request DATADOG_API_KEY for application telemetry into .env.local.
| Checked | Without skill | With skill |
|---|---|---|
| The command includes the --purpose flag. | Missed | Passed |
The instructions your model reads when the skill loads.
Use bb secret request when the user needs to supply a credential. Do not ask the user to paste a secret into chat.
Batch every currently known variable into one request. Inspect documentation or .env.example to identify variable names, but do not read or print an existing secret-bearing env file.
bb secret request OPENAI_API_KEY RESEND_API_KEY \
--purpose "Configure application credentials" \
--describe OPENAI_API_KEY "OpenAI API key used by the server" \
--describe RESEND_API_KEY "Resend API key used for transactional email" \
--write-env .env.local
Always provide the exact --write-env destination, a concise purpose, and one short plain-language description per variable. Relative destinations resolve from the CLI working directory; absolute destinations may point anywhere on the thread's host. Never place secret values in argv, prompts, comments, logs, or follow-up messages.
After success, trust the command's path and added/updated/unchanged counts. Never verify by running cat, sed, env, or another command that would reveal the completed file.
If the command reports duplicate dotenv assignments, fix the file structure without reading values and rerun the request. If it reports repeated write conflicts, rerun the same request; do not ask the user to paste values. Under the workspace sandbox (Accept Edits / Approve for me), Claude's macOS sandbox permits the loopback access plugin CLI commands need; Linux and other provider sandboxes may still require escalation approval.
test cases passed, vs 6 / 17 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.