Refreshes a product’s status using its playbook, projects, tasks, and configured sources. It records observed progress and failures, removes references to archived work, and presents project completion proposals for human review without changing task or project states.
Refresh Product status from current work and source evidence.
Run score Not tested
See how to installThe refresh skill reads a product’s playbook and runs its refresh process. It reviews projects, tasks, and configured sources, then records observed progress and failures in status.md. It narrows the source list and reads relevant documents. If an update conflicts with intervening changes, it rereads the file and incorporates those changes before proposing an update.
The skill removes references to archived work from status.md. It also saves project completion proposals in the wiki and presents them for human review. It does not complete projects, accept work, or mark tasks Ready or Done.
Use it to keep the status of several projects and tasks aligned with progress that can be confirmed in their information sources. It can also help remove references to old work while leaving completion decisions to a person.
The archive period comes from a setting; if none is set, it defaults to 14 days. The skill does not change project or task states. A person reviews and accepts any project completion proposal. If maintenance fails, the skill reports the failure instead of claiming success.
This overview was written by AI (OpenAI Codex) from the public SKILL.md. Check the original below for the exact behavior.
AI checked the SKILL.md for operations that need care and for setup requirements. The score comes only from actually running the skill (hands-on test).
Not counted in the score. Rated by AI (OpenAI Codex) as of Oct 25, 2026. Treat it as a guide and read the original before installing.
Review the instructions the agent will receive before you install.
Read the Product identified in the message. Read its playbook, then call refresh_company_product for that Product. This applies its archive window (metadata.archive_window_days, default 14) and returns Project completion proposals saved in the wiki. It never completes a Project. Present those proposals for a human to accept through the Project transition API; never transition directly. Read the remaining Projects and Tasks, then inspect configured Sources with read_product_source. Reflect observed progress and failures in status.md using its current version. On a conflict, read again and incorporate intervening changes. Propose changes for human review; never accept work or mark a Task Ready or Done. Remove references to archived work from status.md so it no longer enters executor context. Keep brief.md and status.md; maintenance only deletes recognized generated pages older than the window. If nothing changed, return a quiet successful check without writing duplicate pages. Report a maintenance failure instead of claiming success.
For each Source, call read_product_source without path for a paged, content-free index. Use pattern and next_offset to narrow discovery. Pass a returned path to read one document; next_content_offset continues large documents. Compare the full-document content_hash across fragments and reread if it changes. Never infer content from an index entry. Use intake_product_source for Task proposals; propose_company_task is unavailable to Routines.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Reviews code for style, naming, error handling, exposed credentials, and possible bugs, security, or performance issues. It can run an automated style check and returns structured feedback grouped by severity, from critical fixes to optional improvements.
Get practical guidance for Git commits, branch names, and pull requests. The skill follows a consistent commit format, offers naming patterns and a PR description template, and includes common commands for starting, updating, and submitting work.
Reviews configured sources and proposes actionable backlog tasks tied to a suitable project. Each proposal cites inspected evidence and its observed content hash, while stable keys prevent duplicate proposals across later scans.
Summarizes a Product’s recorded changes, pending owner actions, and failures since the last successful delivery. It writes the report to the configured wiki and sends it through the configured interface, or records a quiet check when nothing needs attention.
Inspect the host running the AgentOS process, including its operating system, architecture, Python runtime, processor, and UTC time. You can also list the direct contents of a chosen directory, with the skill folder as the default location.