feat(code-context): ask what each state-changing operation leaves behind (0.2.12)

[alphali]

Summary

Release 0.2.12. Builds on 0.2.11 (#30).

The test-state sweep now requires, for every state-changing operation, one finding per state it leaves behind for the next test:

• what is newly true, or no longer true, afterwards;
• whose state it is: this account, another user, a shared item, or the browser;
• how a tester sees it after a reload or on another screen;
• which operation reverses it, or that it is one-way.

A toast, a loading state or a request name does not count. The findings come from evidence already gathered for the journey, and the agent opens more code only to establish what the operation changes. This keeps the extra cost small.

Why: in live Round DQ-BUF5CB, effects were the weakest dimension, 17/44 on the APOB code-derived benchmark. The findings described what a journey shows, not what it leaves.

Pairs with getdoable/trd claude/code-context-state-left-behind.

Privacy and compatibility

• Skill text only. The helper, manifests and remote data are unchanged, apart from version numbers.
• Compatible with either trd version.

Verification

• npm test passes: release verify 0.2.12, workflow verify, and the 4 node suites.
• No credentials, customer data, private URLs, local paths, or .doable/ state included.
• CHANGELOG, TESTING scenario 26, README and the three manifests are updated.
• claude plugin validate and fresh-session checks were not run.

🤖 Generated with Claude Code