rdd
/rdd
Review an identified code candidate with risk-proportionate lenses and bind findings to that version. Use for receipt-driven review rather than execution testing or implementation.
- Author
- dnlrsls ↗
- Repository
- gentle-stack ↗
- Licence
- MIT
- Last repository change
- 03.11.2026
Assessment and security
- Files in this package
- 3
- Scripts
- 1
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 03.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
- 1 scripts in this package. Review them before installing, as Claude may execute them.
Scripts: scripts/candidate.py
What it does and when to use it
Review an identified code candidate with risk-proportionate lenses and bind findings to that version. Use for receipt-driven review rather than execution testing or implementation.
Original author description
Review an identified code candidate with risk-proportionate lenses and bind findings to that version. Use for receipt-driven review rather than execution testing or implementation.
How it works
- Identify the candidate
- Review
- Receipt
SKILL.md preview
--- name: rdd description: Review an identified code candidate with risk-proportionate lenses and bind findings to that version. Use for receipt-driven review rather than execution testing or implementation. --- # Receipt-Driven Review Adapt Receipt-Driven Development into an on-demand review with an evidence receipt. This skill does not reproduce Gentle AI's native state engine, enforce immutability, or grant delivery authority. ## Identify the candidate Read requirements, relevant Engram decisions, and verification evidence. Record repository, base and head commits, and review scope. For dirty work, capture a diff/content fingerprint including relevant untracked files and baseline; HEAD alone cannot identify uncommitted changes. Use an available read-only snapshot or isolated view when practical. Otherwise compare candidate identity before and after review. If identity cannot be established, disclose that the review is not version-bound. For a reproducible Git candidate, use the optional bundled helper: `python <skill-directory>/scripts/candidate.py --repo <project-directory> --base <review-base>`. Resolve the base once and reuse its returned full commit ID for later comparisons. See [candidate limitations](references/candidate.md) when invoking the helper or choosing an equivalent tool. Python 3 and Git are helper dependencies, not requirements for reading this skill. ## Review Choose relevant lenses based on concrete risks and the requested scope: - **Risk:** security, privacy, data integrity, compatibility, and side effects. - **Resilience:** failures, recovery, retries, concurrency, and partial completion. - **Readability:** unclear contracts or structure that creates maintenance defects. - **Reliability:** behavior, invariants, test relevance, and operational assumptions. Do not require every lens for every change. Cite actionable findings with location, triggering scenario, consequence, and severity. Separate defects from uncertainties and preferences. Passing tests do not prove every reviewed claim. Keep review read-only unless corrections are authorized. If they are, bound them to findings, identify the new candidate, rerun affected checks, and review changed surfaces. Stop at resolved findings, documented remaining issues, or a needed decision; avoid unbounded correction cycles. ## Receipt Return candidate identity, scope and lenses, evidence, findings, gaps, and reviewer provenance. Distinguish self-review from actual independent review. Recheck candidate identity; mark affected conclusions stale and reassess if it changed. No findings means none found within the stated scope, not guaranteed correctness. Consume the task/ticket and acceptance criteria already established by ODD and the checks already observed by verification. Keep evidence tied to its actual candidate and environment. A receipt distinguishes acceptance gaps, actionable findings, and checks not performed. Return findings to the affected unit; a correction requires rechecking changed surfaces rather than restarting all skills. Pass candidate, receipt, unresolved findings, and next action directly to delivery; do not create a mandatory review file or close a ticket merely because review finished. Discover Engram operations and save the receipt summary and references without requiring a repository file. Disclose unavailable persistence and return the unsaved receipt. The receipt does not authorize commit, merge, release, or deployment. For composed work, reuse retrieved context and return the receipt to the coordinating memory owner for a consolidated checkpoint. Standalone saves resolve the workspace/project and use only a bound or successfully registered session if required. Read back uncertain writes before retrying, follow returned conflict handling, and never enable cloud for local memory.
Files in this package (3)
- SKILL.md 4 KB
- references/candidate.md 3 KB
- scripts/candidate.py 3 KB
Versions
The database has no other skills named rdd.
Database snapshot: 13.11.2026 · package version f46d74e93aef (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 90ce-code-review
Review a named diff or PR for bugs, regressions, tests, and standards. Use when asked to review code or when a shipping skill needs a review receipt. Use when asked to apply this review's findings locally. Use ce-resolve-pr-feedback for feedback already left on a PR. - 100code-review
Run an evidence-driven code review that selects correctness, regression, architecture, risk, security, privacy, testing, and performance lenses by change risk. Use for /code-review and general pull-request review requests. - 95review-pr
Review a pull request through multiple quality lenses and present a compiled analysis with inline comments. Use when the user wants a thorough PR review. - 90audit
Use when asked to audit or review a code change, draft plan, or explicitly scoped module, system, or repository. - 90risk-review
Perform a read-only, evidence-based review with correctness always active and additional lenses selected by risk or explicit area. Use after changes or on an explicit review request; do not use to fix findings or replace verification. - 100causal-rdd
Implements sharp and fuzzy regression discontinuity designs in R or Python with bandwidth selection, manipulation testing, and sensitivity analysis. Use when user mentions RDD, cutoff, threshold, running variable, or discontinuity. Not for arbitrary subgroup comparisons.