Ashley Mensah 6d118d9c99 feat(ci): trust LLM decisions and feed it PR merge status ...

- Remove pre_score override from enforcePolicy — policy now only gates
  AUTO_CLOSE, otherwise trusts the model's decision
- Pass pre_score evidence (hard signals, contradictions) to LLM as
  context instead of using it as a decision override
- Fetch linked PR merge status (MERGED/OPEN/CLOSED) in fetch step
  and include in LLM prompt so it can distinguish merged fixes from
  open proposals

2026-04-29 15:49:09 +02:00

..

prompts

fix(ci): distinguish workarounds from actual fixes in system prompt

2026-04-28 17:48:43 +02:00

schemas

fix(ci): add additionalProperties to nested schema objects

2026-04-28 16:11:57 +02:00

scripts

feat(ci): trust LLM decisions and feed it PR merge status

2026-04-29 15:49:09 +02:00

package.json

fix(ci): fix workflow by adding working-directory, fetch script, and removing npm ci

2026-04-28 15:53:17 +02:00