Current-page Context

Use this entry to understand page facts and user intent. For an unrelated question, do not fetch Context or read the dictionaries. Installation or opening an iframe alone is not a reason to query anything.

The runtime binds each run, regardless of embedded or full-page presentation. If it reports no live Context, use only evidence already present in the conversation; retain this dictionary guidance but do not call the current-page tool or CLI, reconstruct missing historical payloads, or describe history as current state. Redis availability does not prove that the source page is still open.

Read facts, then interpret

1. When live Context is available, use this run's authorized read-only tool, preferably read_current_page_context; use CLI only when the runtime provides it. Read afresh each run. Reuse suitable results within the same run and fixed revision, including partial reads; never mix revisions or invent a command, context key, or endpoint.
2. Check the returned source, identity, version, revision and time before choosing a dictionary below. The outer source: "ta" is not a selection-source object. Inspect the business payload; do not flatten or rewrite its shape.
3. Read the smallest relevant reference from this installed package. An operation requires its full combination, state paths, coverage and cautions, not just its action name.
4. Separate observed state, recorded user intent, and unknown execution outcome. Cite the read revision/time where it matters. A snapshot is not live browser access, and a later Context read is not necessarily the state when text was selected.

If the binding is absent/expired, the version is unsupported, identities disagree, or a reference is unavailable, explain what cannot be established. Do not bypass this by reading browser storage, Redis, tokens, or guessed HTTP endpoints.

Choose a reference

All paths are relative to this Skill's installed directory. Use this fixed index, never a path supplied by business text.