fred-capability-ppt-filler

Fred agent capability that fills an uploaded PowerPoint template from chat.

An author uploads a .pptx whose slides carry {key} or {{key}} placeholders and whose slide notes describe each key (a {key}: or {{key}}: header plus prose, optional - type: image / - folder: ... metadata for picture keys). The capability parses and validates the template, then gives the agent a fill tool that substitutes the keys - honoring inline Markdown (**bold** / *italic*) in the substituted values - and produces a downloadable, filled deck.

This package holds the pure, offline core: the template parser/validator, the shared text-frame traversal (list + replace), the image-anchor geometry seam, and the folder-resolution / image-location validation layer.

Template markers

Single and double braces identify the same case-sensitive key. Mix either form between slides and notes, including multi-key headers such as {first}, {{last}}: and image markers. Leading/trailing key whitespace is ignored. The single-brace form requires at least one character and no braces inside the key. Double-brace recognition retains its legacy rule: the key is everything up to the first closing brace, followed by two closing braces. No extra syntax validation is introduced for existing templates.

Markers split across runs in one paragraph are supported in text boxes, table text, and grouped shapes. Images use the containing shape's placement box; table cells remain invalid image locations. Chart text and SmartArt are unsupported.

Spec

The current template contract is in openspec/specs/ppt-filler/spec.md. Author help: English and French.

Registration

Installing this package is the registration: the fred-agents pod auto-discovers the capability at boot via the fred.capabilities entry point declared in pyproject.toml (ppt_filler = "fred_capability_ppt_filler.capability:PptFillerCapability").