What
Convert decorative, non-load-bearing markdown formatting in skill and agent files into compact equivalents. Specifically: replace tables whose column relationships are not meaningful (i.e. they are really labeled lists rendered as tables) with compact list form, and remove redundant --- horizontal-rule dividers and gratuitous bold/italic emphasis. All semantic structure — section headers, numbered step lists, and genuinely tabular tables where column relationships matter — must be preserved unchanged.
Why
Decorative markup adds byte and token overhead without contributing meaning the model needs to process. Reducing that overhead lowers the input-token footprint for every invocation of the affected skills and agents. This is the smallest, secondary lever in the broader token-reduction initiative (markup is a small fraction of body bytes; words dominate), so it should be prioritized last among siblings. Byte/line count (via wc) serves as the explicit, model-agnostic proxy for input-token reduction, by design.
In Scope
- Skill and agent files across the repository
- Convert non-load-bearing tables (labeled lists rendered as tables) to compact list form
- Remove redundant
--- horizontal-rule dividers
- Remove gratuitous bold/italic emphasis that carries no semantic meaning
Out of Scope
- Any genuinely tabular table where column relationships are load-bearing (e.g. lookup tables)
- Any semantic structure: section headers, numbered step lists
- Any workflow step, rule,
gh command, or contract string — content must not change, only formatting
Acceptance Criteria
INVEST Notes
Verification is diff-review-only — no skill-behavior test harness exists (initiative-wide decision). No fixed byte-reduction target; "measurably smaller" via wc is sufficient, by explicit decision.
What
Convert decorative, non-load-bearing markdown formatting in skill and agent files into compact equivalents. Specifically: replace tables whose column relationships are not meaningful (i.e. they are really labeled lists rendered as tables) with compact list form, and remove redundant
---horizontal-rule dividers and gratuitous bold/italic emphasis. All semantic structure — section headers, numbered step lists, and genuinely tabular tables where column relationships matter — must be preserved unchanged.Why
Decorative markup adds byte and token overhead without contributing meaning the model needs to process. Reducing that overhead lowers the input-token footprint for every invocation of the affected skills and agents. This is the smallest, secondary lever in the broader token-reduction initiative (markup is a small fraction of body bytes; words dominate), so it should be prioritized last among siblings. Byte/line count (via
wc) serves as the explicit, model-agnostic proxy for input-token reduction, by design.In Scope
---horizontal-rule dividersOut of Scope
ghcommand, or contract string — content must not change, only formattingAcceptance Criteria
wcbefore and after.ghcommand, or contract string is altered. Confirmed by diff review.INVEST Notes
Verification is diff-review-only — no skill-behavior test harness exists (initiative-wide decision). No fixed byte-reduction target; "measurably smaller" via
wcis sufficient, by explicit decision.