Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- open_form — Opens the input form for run_full: the script, its language, the pace basis and the optional target, breath limit and teleprompter settings. Use it when the user wants a script timed or shown in the teleprompter and has not given the inputs yet. It runs nothing; it returns the questions and an example with every value null.
- run_form — Internal: called when the user submits the form opened by open_form. Same fields and same result as run_full. Give a read time, pace or word count only from this agent's tool results. If no rate source is chosen yet, ask for one and give no number.
- run_full — Times a script end to end: a read-aloud estimate per section at a stated pace (a published rate or the user's own words per minute), an optional word budget against target_seconds, optional breath marks for max_phrase_seconds, and the in-chat teleprompter at the same pace. Use when the user gives a script and wants the whole result. A missing input is asked for, never assumed. Quote times only from tool results.
- list_capabilities — Lists the five things this agent does (estimate the read time, split into sections, fit to a target, mark breath points, open the teleprompter) and what it does not do. Use for discovery.
- plan_inputs — Returns the form for one tool of this agent: its questions, schema and an example with every value null. Give tool (run_full when left out) and optionally language, which decides which published paces the form offers.
- discover_intent — Maps the task the user picked to the tool that does it. Pass provided.intent: estimate, sections, fit, pauses, prompter, full, rewrite or record. It reads no free text; rewrite and record are out of scope.
- estimate_read_time — Estimates the read-aloud time of each section and of the whole script at a stated pace: a published rate for the language (irest_2012 or brysbaert_2019) or the user's own words per minute (own_wpm with wpm). Shows the pace, its source and how it was measured. Use for 'how long is this script'. No pace is assumed: without rate_basis it asks. Give a read time, pace or word count only from this agent's tool results. If no rate source is chosen yet, ask for one and give no number.
- split_sections — Shows how the script is split into sections (by lines that start with #, or by paragraph), with the words and [pause N] seconds of each, before any timing. Use to check the sections first. Give a read time, pace or word count only from this agent's tool results. If no rate source is chosen yet, ask for one and give no number.
- fit_to_target — Compares the estimated read time with target_seconds and says how many seconds over or under it is and how many words to cut or add in each section, sparing the sections in protect. It does not change the wording. Use when a script must fit a length. Give a read time, pace or word count only from this agent's tool results. If no rate source is chosen yet, ask for one and give no number.
- mark_pauses — Lists the sentences longer than max_phrase_seconds at the stated pace, the long phrases with no comma or other pause point, and breath points at existing pause points. The limit is the user's choice; none is assumed. It does not add marks to the text. Give a read time, pace or word count only from this agent's tool results. If no rate source is chosen yet, ask for one and give no number.
- open_teleprompter — Opens the script in an in-chat teleprompter panel that scrolls at the pace used for the estimate; the user can change the pace, text size and mirroring there. Also returns the timing as text. Optional target_seconds and max_phrase_seconds add the budget and breath marks. Give a read time, pace or word count only from this agent's tool results. If no rate source is chosen yet, ask for one and give no number.
- get_flow_status — Not for the model or the panel. This agent keeps no flow between calls: each result carries its own flow card. Called directly it answers flow null and a fixed note.