aevatar-agent-framework

Govern canon changes: require explicit reason/impact/retroactive policy, and produce an actionable propagation plan.

Turn "AI draft -> author edits" diffs into structured deviation items, an author-intent prompt, and a concrete propagation plan.

Help the author write narrative unit tests (rules/constraints) and interpret failures with a concrete fix roadmap.

Simulate multiple reader personas to score a chapter (爽点/困惑/信息负荷) and propose actionable alternatives.

Manage rewrite branches: when to branch, how to compare, and how to merge back with minimal chaos.

Maintain a structured Setup/Payoff ledger (promises, debts, payoffs) and generate a payoff roadmap.

Define what the Story Control Panel shows, how to compute signals, and how the author should act on alerts.

Aggregate a writing session into a readable log + a reusable author-intent prompt.

Evaluate arithmetic expressions precisely by calling py_calc (python-file tool).

Lookup up-to-date library docs via Context7 MCP (resolve-library-id / get-library-docs).

Read environment variables via get_env tool (mask secrets).

Read a small snippet of a text file using file_read.

Grep the workspace for a pattern using file_search.

Pretty-print JSON via json_prettify tool.

Browse local workspace files via MCP filesystem tools (docker-fs).

Convert text to URL-friendly slug via slugify tool.

Compute text stats via text_stats tool.

Provide accurate current time by calling get_time tool.

Provide accurate current time by calling get_time tool.