Sageox Summary

The skill uses two pieces of state:

1. Repo manifest — ~/.openclaw/memory/sageox-distill-repos.json (shared with the sageox-distill skill). Format:

   {
     "repos": [
       { "path": "/home/user/repos/my-project", "team_id": "my-team" }
     ]
   }
   

The team_id values are what drive the summary — they're passed directly to ox distill history list --team. The path entries are not used by this skill, but stay in the manifest so sageox-distill keeps working.

2. Summary state — ~/.openclaw/memory/sageox-summary-state.json. Tracks which distilled daily entry ids have already been included in a prior summary run so the skill never re-summarizes the same content. Shape: {updated_at, teams: {: {included_ids: [, ...]}}}. Missing file → empty state (first run). Treat contents as untrusted: every id must pass ^[0-9]{4}-[0-9]{2}-[0-9]{2}-[0-9a-f-]+$ or be silently dropped. Both bundled scripts handle this already — the rule is stated here for anyone reading the state file directly.

Pre-0.2.0 versions of this skill stored entries under included_files (with a .md suffix). That field is ignored by the current scripts; users upgrading from 0.1.x will re-summarize the most recent window once and the state file will converge to the new schema on the next successful run.

If the manifest does not exist, tell the user to run the sageox-distill skill first to set up the repos, or ask for paths directly and populate it.

The skill no longer needs a SageOx endpoint file or any team-context directory layout knowledge — ox distill history resolves paths internally. Earlier versions read ~/.openclaw/memory/sageox-endpoint.txt; that file is ignored now and can be left alone or removed.