hashgraph-online/awesome-codex-plugins / plugins/boshu2/agentops/skills/research
research
VerifiedAnswer one cited question on code, history, claims or patterns. Use when: asked how it works, why it changed, if a claim or done/shipped report holds, or a pattern warrants a rule.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 17.4 KB |
| references/adoption-and-lineage.md | markdown | 2.4 KB |
| references/codebase-recon/codebase-recon.feature | other | 1.8 KB |
| references/codebase-recon/pack-contract.md | markdown | 3.3 KB |
| references/goals.md | markdown | 1.2 KB |
| references/pattern-mining/pack-contract.md | markdown | 2.2 KB |
| references/pattern-mining/pattern-mining.feature | other | 824 B |
| references/research.feature | other | 530 B |
| references/status.md | markdown | 2.0 KB |
| schemas/findings.json | config | 3.3 KB |
| schemas/reality-check-report.v1.schema.json | config | 1.5 KB |
| scripts/claim-audit/validate-output.shexecutable | script | 1.6 KB |
| scripts/codebase-recon/validate-output.shexecutable | script | 19.0 KB |
| scripts/pattern-mining/validate-output.shexecutable | script | 2.1 KB |
| scripts/validate.shexecutable | script | 2.9 KB |
references/codebase-recon/pack-contract.md
View on GitHub# Codebase-recon pack contract
Applies only when a durable `codebase-recon.v1` pack is selected. An inline
investigation may use dirty working-tree evidence with explicit limits; this
commit-bound pack may not. `<skill-dir>` is defined in [SKILL.md](../../SKILL.md).
A baseline `codebase-recon.json` has this shape (`<...>` is a value to fill):
```json
{
"schema_version": "codebase-recon.v1",
"mode": "baseline",
"commit": "<full commit OID>",
"flows": [{"entry": "<file>", "domain": "<path>", "integration": "<path>", "tests": "<file>"}],
"claims": [{"kind": "fact", "text": "<claim>", "confidence": "high", "evidence": ["<file>:<line>"]}],
"coverage": {"inspected": ["<path>"], "uninspected": ["<path>"]},
"report": {"path": "codebase-recon.md", "sha256": "<lowercase SHA-256 of the companion>"}
}
```
`kind` is `fact`, `inference` or `unknown`; `confidence` is `high`, `medium`
or `low`; both coverage lists are non-empty. A delta manifest sets
`mode: "delta"` and adds `prior_recon` (the prior manifest's path),
`baseline_verified: true` and `delta`, one `{"path": ..., "change": ...}`
entry per changed path; its `flows` may be empty.
- Save `codebase-recon.json` and a cited `codebase-recon.md` companion by
Research's [save rule](../../SKILL.md#output-and-boundaries), outside the
target repository or under its `.agents/` (a new file elsewhere in it is
dirty source; `.agents/scratch/` is disposable). Keep mental model, bounded audit,
pattern evidence and synthesis distinct.
- Bind the exact current full commit OID (not `HEAD` or a short SHA), at least
one complete baseline flow, typed claims and inspected/uninspected scope.
Fact and inference citations, and each flow's `entry` and `tests`, resolve to
repository-relative regular files at that commit (an optional `:<line>` must
exist in that blob); `domain` and `integration` may be directories. The
companion report includes line references. Unknowns remain explicit and may
cite `evidence: []`.
- The manifest `report` names the companion and its lowercase SHA-256. The
companion has one `<!-- codebase-recon-report.v1 -->` marker and
`manifest_commit`, `manifest_mode`, `flows_sha256`, `claims_sha256`, and
`coverage_sha256` markers; section digests hash the `jq -cS` output for each
section, including its trailing newline.
- Discover validated priors with
`bash "<skill-dir>/scripts/codebase-recon/validate-output.sh" --repo-root <target> --discover-priors`;
it scans only `.agents/recon/` and `.agents/scratch/codebase-recon/`, so cite
a prior saved elsewhere by its path in `prior_recon`.
Prefer a verified delta when it answers the request. Delta evidence needs a
valid ancestor chain, `baseline_verified: true` and the exact changed paths
between the prior and current commits; do not relabel a directory scan as delta.
- Run `bash "<skill-dir>/scripts/codebase-recon/validate-output.sh" --repo-root
<target> <recon.json>` before handoff. It checks both artifacts and rechecks
their identities, HEAD and source state; dirty source outside `.agents/` cannot
satisfy this commit-bound pack. Return a validation failure without disguising
it as a completed recon pack.
Preserve earlier packs and their exact cited identities; never move or delete
old proof to match a new layout. See the [recon scenarios](codebase-recon.feature).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 114k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 114k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 114k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 114k | Updated | Claude CodeCodex | Verified | Install |