A Codex skill for adaptive public-internet research.
Dynamic Search treats research as an iterative loop: define the question, plan the evidence needed, search in waves, track source quality, update hypotheses, and synthesize only after uncertainty and role-specific implications are clear.
Use it for unfamiliar, ambiguous, or fast-moving topics where the search strategy should evolve as evidence appears:
- company, market, policy, or technical landscape research
- claim and source audits
- competitor mapping
- public-web diligence
- GitHub, Reddit, forum, and media scanning
- source-backed briefing memos
The skill creates a file-backed research pack with:
- search plan
- source registry
- hypothesis ledger
- evidence matrix
- open questions
- role-perspective map
- final answer or report
- retrospective
Copy this repository into your Codex skills directory, or copy the SKILL.md, references/, scripts/, and agents/ files into a local skill folder named dynamic-search.
The skill entrypoint is:
SKILL.md
The helper script is:
scripts/search_pack.py
Initialize a research pack:
python scripts/search_pack.py init --topic "topic name" --project-dir .Resume and validate an existing pack:
python scripts/search_pack.py migrate --project-dir .
python scripts/search_pack.py validate --project-dir .
python scripts/search_pack.py status --project-dir .See SKILL.md for the full operating rules.