Skill Claude CodeCodex
Trigger when the assistant begins generating explanatory text instead of providing a direct answer in multiple-choice scenarios, ensuring responses match the expected format.
53 2mo ago A 31 tokens
original MIT
Skill Claude CodeCodex
When inferring character intent or behavioral patterns from video, validate inferences against cumulative observed actions rather than isolated moments or assumptions.
53 2mo ago A 31 tokens
original MIT
Skill Claude CodeCodex
When asked to identify critical actions in a video, distinguish between preparatory/support actions and core actions that directly contribute to the main objective, and rank them by significance to the end goal.
53 2mo ago A 44 tokens
original MIT
Skill Claude CodeCodex
When faced with multiple-choice video comprehension questions, systematically eliminate implausible options by cross-referencing visual content, action sequences, and contextual clues before selecting an answer.
53 2mo ago A 40 tokens
original MIT
Skill Claude CodeCodex
When selecting from multiple-choice options about video content, perform exact matching between the question focus and the option text rather than selecting based on partial keyword overlap.
53 2mo ago A 37 tokens
original MIT
Skill Claude CodeCodex
When uncertain between options, systematically eliminate choices that contain factually incorrect elements about the video content, even if other parts seem plausible.
53 2mo ago A 33 tokens
original MIT
Skill Claude CodeCodex
When multiple options describe similar concepts, distinguish between them by comparing precise semantic details (verb choice, object specificity, frequency markers) against video evidence.
53 2mo ago A 35 tokens
original MIT
Skill Claude CodeCodex
When answering multiple-choice questions about video content, systematically identify the PRIMARY action or core process shown, then eliminate distractors that include secondary details or deviations.
53 2mo ago A 35 tokens
original MIT
Skill Claude CodeCodex
When answering questions about video content, ensure the selected answer reflects the PRIMARY OBJECTIVE or OVERARCHING PURPOSE of the actions shown, not just isolated details or secondary effects.
53 2mo ago A 40 tokens
original MIT
Skill Claude CodeCodex
When answering video comprehension questions, explicitly reference observed actions and visual evidence from the video before selecting an answer option.
53 2mo ago A 27 tokens
original MIT
Skill Claude CodeCodex
Trigger when analyzing multi-frame video sequences or action sequences to ensure the assistant correctly identifies the complete sequence of actions and their order rather than selecting incomplete or partially correct options.
53 2mo ago A 40 tokens
original MIT
Skill Claude CodeCodex
Trigger when multiple-choice answers contain similar but subtly different details to prevent selecting plausible-sounding options that don't match the actual visual evidence.
53 2mo ago A 31 tokens
original MIT
Skill Claude CodeCodex
Trigger when output must cite evidence — every factual claim or value must trace to a cited source; never fabricate unsupported values.
53 2mo ago A 30 tokens
original MIT
Skill Claude CodeCodex
Trigger on execcheck / file-production rounds — before finishing, audit the produced file against every explicit requirement stated in the task.
53 2mo ago A 33 tokens
original MIT
Skill Claude CodeCodex
Trigger when a round specifies a schema or file format — match the exact keys, names, value formats and output path; do not add, drop, or rename fields.
53 2mo ago A 40 tokens
original MIT
Skill Claude CodeCodex
Trigger when earlier-round files exist in the workspace — make new output consistent with them, reconciling rather than contradicting.
53 2mo ago A 31 tokens
original MIT
Skill Claude CodeCodex
Trigger on post-update rounds — apply only the changes the update explicitly states; preserve every prior decision the update does not touch.
53 2mo ago A 31 tokens
original MIT