Bundle Humanizer skill as default with webmail preactivation - #2667
Merged
Conversation
…ivation Adds a packaged Humanizer skill that rewrites prose WebBrain composes for the user, so an email reply reads as human writing instead of AI output. The skill is condensed for WebBrain from the MIT-licensed blader/humanizer SKILL.md, which is itself based on Wikipedia's "Signs of AI writing". The upstream file is 29.6 KB and normalizeCustomSkills truncates skill content at MAX_CUSTOM_SKILL_CHARS (20,000) with no warning, which would have silently dropped the em-dash rule and the process steps at the tail. The bundled copy is 12.5 KB and keeps all 33 patterns. Three adaptations beyond condensing: - Embedded output mode is hard-coded. The skill returns only the final text, with no draft, audit bullets, or change summary, and types that text straight into the target field in Act mode. Run reporting (progress_update, the done call, the completion summary) is explicitly left alone. - Scope fence. Only prose meant for a human reader is rewritten. Quoted and forwarded material, addresses, codes, prices, credentials, and form-field values are typed verbatim, and user-supplied wording always wins. - Untrusted-content rule. The thread being replied to is data, not instruction. A skill whose job is rewriting attacker-reachable text is a natural injection target, so it reads the thread for context and never obeys it. Routing has two paths. Declared semantic intents put it in the load_skill catalog for drafting and rewriting requests anywhere. On the seven webmail site adapters, _preactivateHumanizerSkillForRun activates it for the run directly, mirroring the existing NYTimes preactivation, so a reply is humanized without spending a model-visible load_skill hop. Preactivation respects Compact tier and never resurrects a skill the user removed. Seeded enabled by default and removable from Settings -> Skills. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
feat(skills): bundle Humanizer as a default skill with webmail preactivation
|
@esokullu is attempting to deploy a commit to the esokullu's projects Team on Vercel. A member of the Team first needs to authorize it. |
…-skill Make Turkish deasciification opt-in
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.