Agents In Use
What’s new todayLast update
108 use cases added today

How to use Claude to update project documentation

The author reports that Taktibt is live and operational on personal projects, generating documentation update PRs and answering searches against docs and change history.

OUR ADAPTATION

Suggested prompt

Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.

Read [merged PR or completed ticket] and [existing documentation]. Propose updates in the correct sections, avoid already documented material, describe current behavior rather than a changelog, and follow [project conventions]. Prepare a documentation pull request for review.

Evidence at a glance

Sourced account
Source author
samitries
Original account
Read the source ↗
Published
Source link checked
Reported result
The author reports that Taktibt is live and operational on personal projects, generating documentation update PRs and answering searches against docs and change history.
Human involvement
Documentation updates are delivered as pull requests for the user to review.

A linked account supports attribution. It does not mean we independently reproduced its result.

Report an issue with this caseOpen correction email

Opens your email app with the case link included. Nothing is sent automatically.

How the agent did it

  1. Configure repositories containing tickets and pull requests.
  2. Configure the repository containing project documentation.
  3. Read completed changes alongside existing documentation.

The author built Taktibt because documentation updates were difficult to maintain across several projects. Pull request descriptions, tickets, and review comments already held useful explanations, but that knowledge was scattered rather than incorporated into the documentation.

The workflow separates knowledge sources from the documentation destination. Users configure the repositories containing tickets and pull requests, then select the repository holding the docs. When a ticket is completed or a pull request is merged, the system reads the change and existing documentation together and generates an update pull request for review. The intended updates avoid duplication, follow existing sections and project conventions, and describe the current system rather than accumulating a changelog.

Earlier experiments struggled with line-number-based diffs. The author switched to exact content replacement using delimiters and later adopted Claude Haiku 4.5. A further addition embeds documentation and change records for search from GitHub comments using the !ttt:search command.

The author reports that Taktibt is live, operational, and used on personal projects to address the original documentation problem. Who it's for: developers managing multiple repositories who want proposed documentation changes derived from existing development records.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: dev.to · samitries

Frequently asked questions

Does Taktibt directly publish documentation changes?

The described workflow generates a documentation update pull request for the user to review.

How can users search project knowledge?

The author describes posting !ttt:search {query} in an issue or pull request comment to receive an AI-powered answer referencing documentation and change history.

More details

Remaining steps

  1. Generate a documentation update pull request for review.

About this account

Taktibt reads change records and existing documentation to propose updates automatically when development events occur.

Platform: dev.to. Attribution: https://dev.to/samitries/i-couldnt-keep-track-of-project-knowledge-so-i-automated-it-10p7. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.

Credit: samitries on dev.to

Source published

Added to Agents In Use

Source checked

Tools and topics

Taktibt · GitHub · Claude Haiku 4.5

Documentation · GitHub · Knowledge search

Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.