When dogfooding processkit v0.27.1 in projectious-work/aibox, pk-doctor reports:
AGENTS.md managed block 'pk-commands' differs from context/templates/processkit/v0.27.1/AGENTS.md
The local AGENTS.md command adapter intentionally contains the consuming project's commands (cargo build/test/clippy/fmt for aibox). The upstream processkit template contains processkit's own commands (npm/uv smoke). Comparing the block byte-for-byte against the upstream template makes every correctly customized consuming project look drifted.
Expected: pk-doctor should validate the pk-commands schema and markers, but should not require command values to match the processkit repository template unless a project explicitly opts into that policy.
Observed with processkit v0.27.1 during aibox v0.27.3 integration.
When dogfooding processkit v0.27.1 in projectious-work/aibox, pk-doctor reports:
AGENTS.md managed block 'pk-commands' differs from context/templates/processkit/v0.27.1/AGENTS.md
The local AGENTS.md command adapter intentionally contains the consuming project's commands (cargo build/test/clippy/fmt for aibox). The upstream processkit template contains processkit's own commands (npm/uv smoke). Comparing the block byte-for-byte against the upstream template makes every correctly customized consuming project look drifted.
Expected: pk-doctor should validate the pk-commands schema and markers, but should not require command values to match the processkit repository template unless a project explicitly opts into that policy.
Observed with processkit v0.27.1 during aibox v0.27.3 integration.