v0.0.32 — token-efficient AI callers

Highlights

Release theme: token-efficient AI callers. Two closely related shifts. First, response payloads shrink dramatically for the common "small edit" and "just tell me the block structure" intents. Second, ability descriptions gain author-declared hints pointing AI callers at cheaper sibling abilities when their intent maps to one. Every change is either strictly additive or opt-out via an admin toggle — no ability's execute() behaviour changes.

What's in 0.0.32

• PR #152 — return_content:false default on six content/{create,update}-{page,post,cpt-item} writers. Single-word edit on a ~97 KB page dropped from ~60K → ~29K LLM tokens. BREAKING for callers reading response.page.post_content etc. — pass return_content:true explicitly.
• PR #153 — Same default on nine blocks/{create,update}-* writers: patterns, templates, template parts, style variations, global styles. For Variation_Db / Global_Styles_Db writers, the caller's $return_content flag is now passed to to_row() so decode_content() is skipped when the payload isn't wanted (CPU savings on the hot path).
• PR #154 — New blocks/outline-post-blocks ability. Flat, depth-first content-free block index — canonical path, block type, child count, byte size, short text preview. ~28K tokens cheaper than get-post-blocks on the same page. Path convention matches Block_Tree (raw parse_blocks() indices, whitespace nodes counted) so paths are drop-in usable with blocks/add-block / blocks/update-post-block / blocks/remove-block. Response includes post_modified_gmt for staleness detection.
• PR #155 — Feature 095 suggested_abilities() framework. New protected method on Ability_Definition for declaring cheaper alternative abilities (~10-30× token savings pointer). Entries surface under meta.acrossai.suggested_abilities on mcp-adapter-get-ability-info. New admin toggle "Disable ability suggestions" strips the field site-wide (option acrossai_disable_ability_suggestions, default 0 = enabled). Ships with 4 initial overrides on content/update-{page,post,cpt-item} + blocks/get-post-blocks.
• PR #156 — Ten more suggested_abilities() overrides on the hint catalog:
  • content/get-page / get-post / get-cpt-item → blocks/outline-post-blocks (~20×)
  • blocks/read-theme-json → blocks/get-style-guide (~5–8×)
  • options/list-options → options/get-option (~15–30×)
  • media/list-media → media/get-media (~10–15×)
  • users/list-users → users/get-user (~10–20×)
  • blocks/list-block-templates → blocks/read-block-template (~5–15×)
  • blocks/list-block-patterns → blocks/read-block-pattern (~8×)
  • blocks/list-global-styles → blocks/read-global-style (~10×)

Test status

• Full suite 2165 / 2165 passing (+94 vs 0.0.31)
• PHPCS + PHPStan level 8 clean on every changed file
• Live MCP smoke verified for content/update-page, blocks/create-block-pattern, blocks/create-block-style-variation, blocks/outline-post-blocks, content/get-page, blocks/list-global-styles, media/list-media, and the kill-switch toggle on/off

Breaking notes for existing callers

• Any caller reading response.page.post_content / .post.post_content / .item.post_content on the six content writers must now pass return_content:true explicitly.
• Any caller reading response.pattern.content / response.template.content / response.part.content / response.variation.data / response.record.data on the nine block-editor writers must now pass return_content:true explicitly.
• No other callers are affected.

Install

Download acrossai-abilities-manager.zip below and upload via Plugins → Add New → Upload Plugin, or replace the existing wp-content/plugins/acrossai-abilities-manager/ folder.