Skip to content

Execute stage skill #57

Description

@savitharaghunathan

Part of #54.

Execute

Reads PLAN.md and applies transformations file by file. Supports parallel sub-agents.

Consumes PLAN.md, domain skill references (/opt/skills/*/references/)
Produces Migrated source files committed to branch, .konveyor/results.json entry
Image agent-execute-{lang} (e.g. agent-execute-java)
Mode Non-interactive, supports parallel sub-agents

Parallel sub-agents

Each plan step (or group of steps) can run as a separate sub-agent. Each sub-agent handles one step from PLAN.md — reads the target file, applies transformations, writes it back.

Git strategy (TBD): Sub-agents working in parallel need a git merge strategy:

  • Each sub-agent could work on a separate branch and results get merged back into the main migration branch
  • Or sub-agents work on disjoint files on the same branch (PLAN.md steps are one-file-per-step, so conflicts are unlikely if steps are partitioned correctly)
  • Need to handle merge conflicts if two steps touch the same file (e.g. a shared config file)

Expected artifact

Migrated source files committed to the branch. Results entry includes step count and completion summary.

{"stage": "execute", "status": "succeeded", "summary": "migrated 47/47 steps"}

Tasks

  • Design sub-agent per plan step strategy
  • Define git branching/merge strategy for parallel sub-agents
  • Handle merge conflicts when multiple steps touch shared files (e.g. pom.xml, config)
  • Skill-level chunking for context window scaling (process N steps at a time)
  • Define how partial completion is reported (step X of Y completed before context exhaustion)
  • Ensure domain skill references are consumed for transformation patterns

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions