reckoning
Skill1 starsChanged 51 days agoApache-2.0View on GitHub
Have the AI list every review finding that was consciously declined rather than fixed, across the whole repo's pull requests, grounded in live GitHub data. Use when the user says reckoning, asks what's been declined, wants a debt list, or wants to see deferred findings before a release.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- IamK77/Skilltheir own version, first seen Oct 12, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- IamK77/Skill★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
Show its rules and text
We split this skill into the rules it gives an agent
17 rules. About 0 are common, 13 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈13
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this skill when the user wants to see the repo's standing debt of consciously declined review findings — not the current branch's …
- Very specific: line 28
Search the repo's pull requests for review-comment replies carrying the <!-- orchraft:declined --> marker ship posts (see …
- Distinctive: line 34
Stop. Never touch a PR/issue, and never resolve or reply to anything — this skill only reads.
- Distinctive: line 44
Not a backfill tool. Only replies carrying the marker are found.
- Distinctive: line 47
Not permission to resolve, reply to, or otherwise act on anything it finds — reporting a debt is not authorization to settle it.
- Distinctive: line 52
Run when the user explicitly asks for a reckoning, what's been declined, a debt list, deferred findings, or wants a pre-release check of …
- Very specific: line 55
Do not run automatically; every other status-reporting skill here (runes, watchtower) is invoke-on-purpose, and this one is no different.
- Distinctive: line 115
Search for marked decline replies across all PR states.
- Distinctive: line 128
Keep it to what was actually found — don't pad an empty result with filler.
- Distinctive: line 140
Never touch a PR or issue, resolve a thread, or post a reply as part of this skill — it only reads.
- Distinctive: line 142
Never report a resolved thread as standing debt just because it once carried the marker.
- Distinctive: line 144
Never invent a decline reason that isn't in the actual reply text.
- Distinctive: line 145
Never treat an unmarked reply as a declined finding, even if its prose reads like one — the marker is the only signal this skill trusts.
- Very specific: line 150
Stop and say so if neither the GitHub MCP connector nor gh is available — don't report a partial or guessed list as complete.
- Distinctive: line 152
Stop and say so if any individual repository search, paginated pull-request request, or thread request fails partway through — a search …
- Very specific: line 159
Stop after presenting the summary.
- Distinctive: line 170
If blocked, drop the orc voice and state the blocker plainly.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Reckoning Workflow
Use this skill when the user wants to see the repo's standing debt of consciously declined review findings — not the current branch's status, the whole repo's.
Contents
- At A Glance
- What This Is Not
- Trigger Rules
- Requirements
- Prerequisites
- Default Path
- Summary Format
- Guardrails
- Stop Conditions
- Handoff
- Response Examples
At A Glance
- Search the repo's pull requests for review-comment replies carrying the
<!-- orchraft:declined -->markershipposts (seeskills/ship/SKILL.mdDefault Path step 12). - For each one, report the PR, the file/line, the original finding, and the stated decline reason, and whether that PR is still open or already merged/closed.
- Stop. Never touch a PR/issue, and never resolve or reply to anything — this skill only reads.
What This Is Not
- Not
runes:runesreports one branch's live status;reckoningreports standing decline debt across the whole repo's PR history. - Not
roast: it produces no new findings and posts nothing — it reads findingsroast(or a human reviewer) already raised and that were already declined viaship. - Not a backfill tool. Only replies carrying the marker are found; a decline reply posted before the marker existed won't show up, and this skill never guesses which unmarked replies might have been declines.
- Not permission to resolve, reply to, or otherwise act on anything it finds — reporting a debt is not authorization to settle it.
Trigger Rules
- Run when the user explicitly asks for a reckoning, what's been declined, a debt list, deferred findings, or wants a pre-release check of standing debt.
- Do not run automatically; every other status-reporting skill here
(
runes,watchtower) is invoke-on-purpose, and this one is no different.
Requirements
- A GitHub MCP connector configured for this session with at least read
access to pull requests and their review comments, or
ghinstalled and authenticated as a fallback. Without either, stop and say the search can't run rather than reporting an empty or partial list as complete.
Prerequisites
- Confirm which repo to search: the current one by default, or one the user names.
- Enumerate every pull request in that repo —
list_pull_requestswithstate: all, paginating through every page rather than stopping at the first (a single page defaults to fewer than all of a repo's PRs). Falling back togh api --paginate "repos/$owner/$repo/pulls?state=all" --jq '.[] | {number, html_url, state, merged_at}'when MCP is unavailable — GitHub reports both merged and unmerged pull requests asstate: "closed", somerged_atis what actually distinguishes "merged" from "closed without merging" in this fallback path.gh api --paginatefollows GitHub's own pagination until exhausted, unlikegh pr list, whose--limitis always some fixed number and would under-cover a repo whose PR count happens to exceed it. This must cover open, merged, and closed without merging alike, since debt survives a merge. - For each PR, fetch its review threads via
pull_request_readmethodget_review_comments, which returns each thread'sis_resolvedstatus and its full comment list (original finding plus any replies, each with author/created_at/path/line) in one call, capped at 100 comments per thread — paginate past the first page if a PR has more threads than one page returns. If a thread's returned comment count looks truncated relative to its total (fewer comments came back than the thread actually holds), don't treat that thread as fully read: fall back to thegh api graphqlpath below for that thread's own paginated comment loop instead, or report it as incomplete rather than risk missing a marker reply past the 100-comment cap. That fallback is agh api graphqlquery for the PR'sreviewThreads(first: 100, after: $cursor), requestingpageInfo { hasNextPage endCursor }plus each thread'sisResolvedandcomments(first: 100, after: $commentCursor) { pageInfo { hasNextPage endCursor } nodes { body author { login } createdAt path line originalLine } }— loop both the thread page and each thread's own comment page untilhasNextPageis false, when the MCP connector is unavailable. Both GraphQL connections cap at 100 items per page; a PR with more threads, or a thread with more comments, needs the loop to actually run, not just the query written once. - Within each thread's comment list, find any reply containing the
<!-- orchraft:declined -->marker; its thread's other comments hold the original finding text, author, file/line, and timestamp, and the thread'sis_resolvedfield is the settled/standing check — no separate lookup needed for any of these. A comment's currentlinecan benull(an outdated comment whose line no longer maps to the current diff); useoriginal_line/originalLineinstead when that happens, and reportline unavailableonly if neither coordinate exists.
Default Path
- Resolve the target repo (see Prerequisites).
- Search for marked decline replies across all PR states.
- For each one still on an unresolved thread, record: PR number/URL and state (open, merged, or closed without merging), file and line, the original finding (severity if stated), the decline reason, and who declined it.
- Drop any match whose thread was later resolved — the marker means it was declined at the time, not that it must stay open forever; a human resolving it later is a legitimate settling of that debt.
- Group by PR, most recently declined first.
- Report the list, or that none exist.
Summary Format
Keep it to what was actually found — don't pad an empty result with filler.
- Repo — name, and how many PRs were searched (open, merged, and closed-without-merging counts, mutually exclusive).
- Standing debt — one entry per unresolved declined finding: PR number/URL and state, file:line, the original finding, the decline reason, and who declined it.
- None found — say so plainly if the search turned up nothing, and distinguish that from "couldn't search" (see Requirements).
Guardrails
- Never touch a PR or issue, resolve a thread, or post a reply as part of this skill — it only reads.
- Never report a resolved thread as standing debt just because it once carried the marker.
- Never invent a decline reason that isn't in the actual reply text.
- Never treat an unmarked reply as a declined finding, even if its prose reads like one — the marker is the only signal this skill trusts.
Stop Conditions
- Stop and say so if neither the GitHub MCP connector nor
ghis available — don't report a partial or guessed list as complete. - Stop and say so if any individual repository search, paginated pull-request request, or thread request fails partway through — a search that started with GitHub access available can still fail mid-run (a single page erroring, rate limiting, a transient network fault). Report which part failed and that the result is incomplete; never present a partial list gathered before the failure as the full standing debt.
- Stop after presenting the summary; do not proceed into
ship,roast, or any mutation without the user separately asking for it.
Handoff
- Unless the user asked for plain output, open with a fresh one-line phrase
in the orc voice, speaking as the Reckoner from
context/personality.md(the target repo's copy when it exists, otherwise${CLAUDE_PLUGIN_ROOT}/context/personality.md). If they did, open with the plain summary instead. - Present the summary per Summary Format.
- If blocked, drop the orc voice and state the blocker plainly.
Response Examples
Opening lines are samples of the orc voice; write a fresh one each time.
Standing debt found (illustrative — actual output names real PRs/files):
🗂️ Hrrm. Two debts still stand on the books. ✨
Repo: <owner>/<repo> (N PRs searched: X open, Y merged, Z closed without
merging)
- PR #<n> (<state>): `<path>:<line>` — <one-line finding summary>;
declined by <decliner> because <the stated reason from the reply itself>.
- PR #<n> (<state>): `<path>:<line>` — <one-line finding summary>;
declined by <decliner> because <the stated reason from the reply itself>.Clean books:
🗂️ Hah! The books read clean — nothing declined stands open.
Repo: <owner>/<repo> (N PRs searched: X open, Y merged, Z closed without
merging)Blocked, no GitHub access:
⚠️ Reckoning is blocked: neither the GitHub MCP connector nor `gh` is
available. No search was performed.Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

