Autonomous Direction

Seven tools that power the /direct pipeline. Extract briefs from project context, plan story beats across multiple strategies, score and revise candidates, and run automated revision loops until convergence.

---

extract_story_brief

Extract a structured story brief from project context. Analyzes project assets, brand guidelines, and intent to produce a brief with a quality score and warnings for missing information.

Returns: Structured brief with sections (hook, narrative, CTA), quality score (0-100), and warnings for gaps.

Try asking your AI:

"Extract a story brief from the fintech-demo project"

---

plan_story_beats

Convert a brief and archetype into a beat plan with durations, camera moves, and continuity markers. Generates multiple strategies (typically 3) for comparison.

Returns: Beat plan per strategy with scene durations, camera intent, transition types, and continuity IDs.

Try asking your AI:

"Plan 3 story beat strategies for this brief using the crescendo archetype"

---

score_candidate_video

Unified 6-dimension scorecard for a candidate video. Evaluates with per-scene subscores so you can pinpoint exactly where quality drops.

The six dimensions:

• hook - Opening impact and attention capture
• narrative_arc - Story structure and progression
• clarity - Message clarity and visual hierarchy
• visual_hierarchy - Compositional quality per scene
• motion_quality - Animation smoothness and appropriateness
• brand_finish - Brand consistency and polish

Returns: Overall score (0-100), per-dimension scores, per-scene subscores, and specific improvement notes.

Try asking your AI:

"Score this candidate video against the original brief"

---

revise_candidate_video

Apply targeted revisions to a candidate manifest. Nine bounded transforms are available: trim, swap, reorder, retime, retransition, recamera, reinsert, remove, and regrade.

Returns: Revised manifest with applied transforms and a changelog of what changed.

Try asking your AI:

"Trim scenes 3 and 5 to 2 seconds each and swap scenes 4 and 6"

---

compare_candidate_videos

Rank 2-3 candidate videos with trade-off analysis. Compares candidates across all six scoring dimensions and highlights where each candidate excels or falls short.

Returns: Ranked candidates with per-dimension comparison, trade-off analysis, and a recommended winner.

Try asking your AI:

"Compare these 3 candidate videos and recommend the best one"

---

auto_revise_loop

Automated score-revise-rescore loop. Scores the current candidate, applies targeted revisions, re-scores, and repeats until the score converges or a maximum iteration count is reached.

Returns: Final revised manifest, iteration log with scores at each step, and convergence status.

Try asking your AI:

"Run the auto-revise loop on this manifest until it scores above 85"

---

generate_brief_stub

Generate a structured brief markdown file from project context. Useful for bootstrapping a project brief that can be edited by hand before running the full direction pipeline.

Returns: Markdown brief document with sections for hook, narrative, scenes, brand, and delivery.

Try asking your AI:

"Generate a brief stub for a new product launch project"