Research Agent Workflow - joeroberts/release-radar GitHub Wiki

Agent workflow research

This section preserves the agent-workflow research named by issue #163. The dated source records are historical evidence. They are not operating standards, accepted architecture decisions, implementation plans, or authorization to make changes.

Start here

  • Agent workflow research synthesis — 2026-10-02 links the major findings, corrections, conflicts, evidence categories, and unresolved sequencing questions back to exact source sections.
  • Use the verbatim records below when wording, chronology, qualification, or authorization boundaries matter.

Dated source records

Source task Coverage Record
Compare agent skills workflow Two completed turns: initial Anthropic article comparison, then corrected environment-specific review 2026-10-02 source record
Plan workflow skill integration One completed turn: integrated comparison, script probes, effectiveness limits, and proposed sequence 2026-10-02 source record
Compare Cursor pstack skills Three completed turns: initial comparison, progressive-disclosure correction, and global-instruction design investigation 2026-10-02 source record
Compare external skills One completed turn: JavaScript Mastery skill-suite comparison 2026-10-02 source record

Preservation rules for these records

  • The source pages reproduce all completed pre-assignment user and assistant messages in chronological order.
  • The later preservation assignment is the cutoff and is not part of the historical research.
  • Raw tool output is not copied wholesale. Reported revisions, measurements, experiments, source links, qualifications, and selected meaningful evidence output are preserved; truncation and omissions are disclosed.
  • Private runtime instructions, secrets, unrelated task inventories, internal reasoning, and long third-party file copies are excluded.
  • The original Codex tasks remain intact. These pages are durable retrieval records, not replacements for those tasks.