How Promptless protects secrets and sensitive data

Promptless protects secrets and other sensitive data in two layers. Data minimization comes first: Promptless processes trigger and integration content in real time and does not retain it. Automatic credential redaction adds a second layer, replacing credential formats and known secret values in the records Promptless logs, stores, and transmits. That coverage extends to those formats and values only. Promptless does not detect or redact personal data (PII) in the content it processes. Data minimization covers that non-credential sensitive data. This page describes both layers so a security or procurement reviewer can confirm how sensitive data is protected.

How Promptless protects your data

Data minimization is the primary protection. Promptless processes the content behind a trigger or context source in real time to generate a suggestion, and does not store it. This covers a GitHub PR trigger, a Slack thread, and context from tools like Linear, Jira, Confluence, and Notion. Promptless also applies these limits:

• Promptless does not retain a copy of source code, support conversations, or other integration data.
• Promptless does not process sensitive end-user data.
• Promptless uses your content only to generate your documentation suggestions, and does not use it to train language models.

For the full data-handling posture, see data handling and classification. For how this applies to AI generation, see how Promptless uses AI.

Credential redaction

Credential redaction is the defense-in-depth layer alongside data minimization. It is automatic and always applied, with no toggle. It matches credential formats and specific values you mark as secret using environment variables, acting on those recognized patterns and known values only. You mark a value secret with the Secret toggle under Agent Env Vars in Settings. These mechanisms protect the records Promptless logs, stores, and transmits. They do not scan the trigger and context content the agent reads to generate a suggestion.

Redaction applies at these destinations:

• Log previews, telemetry spans, and persisted tool-call records: Promptless replaces GitHub and GitLab token formats and TOKEN=-style assignments with *** or KEY=***.
• Structured request logs: Promptless replaces the values under sensitive key names with [REDACTED] before an inbound request body reaches a structured log. Sensitive key names include password, token, and client_secret.
• Browser-tool responses: Promptless uses its browser automation tool for tasks like capturing screenshots with Promptless Capture. Before the response reaches the agent, it rewrites each value you marked as secret to <secret>NAME</secret>.

What redacted output looks like

The examples below are illustrative and use fabricated placeholder values. Redacted records take one of three forms.

A recognized credential becomes ***, and a TOKEN=-style assignment keeps its key and becomes KEY=***:

ghp_exampletoken            ->  ***
GH_TOKEN=ghp_exampletoken   ->  GH_TOKEN=***

Values under sensitive key names become [REDACTED] in structured logs:

{
  "password": "[REDACTED]",
  "client_secret": "[REDACTED]"
}

A known secret value in a browser-tool response becomes <secret>NAME</secret>, where NAME is the label you gave the value:

<secret>DEPLOY_TOKEN</secret>

What Promptless does not scan

Credential redaction covers the credential formats and known secret values described above. Its scope is limited to those:

• Promptless does not run a general secret or personal-data (PII) detector over ingested content from Confluence, Notion, Jira, Linear, Slack, or GitHub.
• Promptless does not detect Social Security numbers, credit-card numbers, phone numbers, or free-form personal data.
• Promptless does not scan the suggestion body for secrets before publishing.

Personal data present in a connected source, such as a ticket or a chat message, can appear in a generated suggestion. Keeping such data out of the systems Promptless reads from is your responsibility.

Encryption and isolation

For encryption standards and the multi-tenant isolation model, see network architecture and data handling and classification. For how Promptless notifies you of a security incident, see compliance and certifications.