feat(skills): /issue — file GitHub issues from arbitrary input#35
Merged
samtuckerdavis merged 3 commits intomainfrom Apr 26, 2026
Merged
feat(skills): /issue — file GitHub issues from arbitrary input#35samtuckerdavis merged 3 commits intomainfrom
samtuckerdavis merged 3 commits intomainfrom
Conversation
|
Warning Rate limit exceeded
Your organization is not enrolled in usage-based pricing. Contact your admin to enable usage-based pricing to continue reviews beyond the rate limit, or try again in 4 minutes and 40 seconds. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (2)
📝 WalkthroughWalkthroughAdds operator-facing tooling docs: defines Changes
Sequence Diagram(s)sequenceDiagram
participant Operator as Operator
participant IssueSkill as /issue Skill
participant RepoContext as Repo Context
participant Deduper as Deduplicator
participant Sanitizer as Sensitivity Sanitizer
participant GH as gh CLI
Operator->>IssueSkill: invoke `/issue` (input + flags)
IssueSkill->>IssueSkill: startup gate (gh auth status) / refuse if bot identity
IssueSkill->>IssueSkill: parse mode (A/B/C) & repo inference
Note right of IssueSkill: Mode A → single draft\nMode B → decompose candidates
IssueSkill->>Deduper: run dedup classification per candidate
IssueSkill->>Sanitizer: sanitize bodies by repo tier
IssueSkill->>Operator: show approval table / previews
Operator->>IssueSkill: approve selection (or --no-confirm)
IssueSkill->>GH: `gh issue create` for approved items (unless dry-run)
GH-->>IssueSkill: return create result
IssueSkill->>Operator: report success/failure and partial state
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 8✅ Passed checks (8 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.claude/rules/tooling-reference.md:
- Line 98: Remove the embedded calibration detail from the `/merge [--risky]
[--repo <name>]` reference line—specifically drop the phrase describing MED/HIGH
calibration (e.g., "the calibration is built so MED is the median outcome, not
HIGH")—and leave only the concise tool description; move or ensure the full
calibration specifics live only in the `/merge` skill spec (SKILL.md) so this
reference doc stays DRY and minimal.
- Line 97: Update the /run command documentation to specify the exact UTC
timestamp filename format used for reports so downstream tools can parse
filenames reliably; explicitly state the pattern for the placeholder in
"~/.claude/orchestrator-log/run-<UTC-timestamp>.md" (for example an ISO-like,
filename-safe pattern such as YYYY-MM-DDTHHMMSSZ or YYYY-MM-DD_HHMMSSZ) and note
it is UTC, and mention any separators or timezone marker expected so consumers
of the /run report can implement matching parsing logic for the /run report
files.
In `@claude-code/.claude/skills/issue/SKILL.md`:
- Around line 203-239: The sentence under "STEP 4 — Sensitivity sanitization"
that reads "Private Open-Paws repos (if any) → `private` → no strip required,
but never copy verbatim from a confidential origin." is ambiguous; change it to
an explicit clarification so readers know private-tier repos do not require
public redaction but verbatim pasting of confidential transcripts/emails is
still forbidden. Edit that bullet (referencing the exact phrase) to something
like: "Private Open-Paws repos (if any) → `private` → no public-strip required,
but do NOT paste raw/confidential transcripts or emails verbatim — paraphrase or
summarize instead." Ensure the new wording aligns with "Verbatim is forbidden"
and the strict-strip/mild-strip rules in STEP 4.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 897cb9fd-2058-4a99-9880-2d9de1f0e5ae
📒 Files selected for processing (2)
.claude/rules/tooling-reference.mdclaude-code/.claude/skills/issue/SKILL.md
Contributor
Author
|
STAGE 7 implementation complete — addressed all 3 CodeRabbit findings. Pushed b3c53f6. |
Contributor
Author
|
@coderabbitai full review please — all 3 actionable findings addressed in b3c53f6. |
|
🧠 Learnings used✅ Actions performedFull review triggered. |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@claude-code/.claude/skills/issue/SKILL.md`:
- Line 34: Several fenced code blocks in SKILL.md are missing language
specifiers; update each fenced block by adding an appropriate language
identifier (e.g., text or markdown) so screen readers and plain-text editors
handle them correctly. Specifically, find the block containing "/issue refuses
to run under bot identity (active: OpenGaryBot)." and add `text`; update the
blocks that begin with "**Observed:** <what's happening..." and "## /issue —
Mode A (direct file) → Open-Paws/<repo>" and "## /issue — Mode B → extracted N
candidates from input" and the lone "Filed:" and the quoted line starting with
"> from:" to use `markdown` as the fence language. Ensure each triple-backtick
fence is changed from ``` to ```text or ```markdown accordingly.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 204d431f-29af-49c4-9382-d4913b7af57d
📒 Files selected for processing (2)
.claude/rules/tooling-reference.mdclaude-code/.claude/skills/issue/SKILL.md
|
|
||
| Parse the active account. If it is `OpenGaryBot` (or any other bot identity matching `*Bot` / `*-bot`), refuse: | ||
|
|
||
| ``` |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
Add language specifiers to fenced code blocks for accessibility.
Six fenced code blocks lack language specifiers, affecting screen readers and plain-text terminals. Coding guidelines require instruction files to work in editors with no syntax highlighting and screen readers.
♻️ Proposed fixes
Line 34: Add text or markdown identifier
-```
+```text
/issue refuses to run under bot identity (active: OpenGaryBot).Line 89: Add markdown identifier
-```
+```markdown
**Observed:** <what's happening or what the operator noticed; one paragraph max>Line 106: Add markdown identifier
-```
+```markdown
## /issue — Mode A (direct file) → Open-Paws/<repo>Line 155: Add markdown identifier
-```
+```markdown
## /issue — Mode B → extracted N candidates from inputLine 178: Add markdown identifier
-```
+```markdown
Filed:Line 226: Add markdown identifier
-```
+```markdown
> from: <one-sentence paraphrase or sanitized quote of the source>As per coding guidelines: "Instruction files must work in editors with no syntax highlighting, screen readers, and plain-text terminals (Accessibility)".
Also applies to: 89-89, 106-106, 155-155, 178-178, 226-226
🧰 Tools
🪛 markdownlint-cli2 (0.22.1)
[warning] 34-34: Fenced code blocks should have a language specified
(MD040, fenced-code-language)
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@claude-code/.claude/skills/issue/SKILL.md` at line 34, Several fenced code
blocks in SKILL.md are missing language specifiers; update each fenced block by
adding an appropriate language identifier (e.g., text or markdown) so screen
readers and plain-text editors handle them correctly. Specifically, find the
block containing "/issue refuses to run under bot identity (active:
OpenGaryBot)." and add `text`; update the blocks that begin with "**Observed:**
<what's happening..." and "## /issue — Mode A (direct file) → Open-Paws/<repo>"
and "## /issue — Mode B → extracted N candidates from input" and the lone
"Filed:" and the quoted line starting with "> from:" to use `markdown` as the
fence language. Ensure each triple-backtick fence is changed from ``` to ```text
or ```markdown accordingly.
…y input
Operator-only slash command. Takes anything from a one-sentence note to a
pasted meeting transcript and turns it into well-formed GitHub issues that
scout/triage would themselves accept. Auto-detects single-vs-batch — the
operator never declares a mode.
Three modes, deterministic detection:
- Mode A (≤3 sentences + resolvable repo): draft + sanitize + dedup-probe +
confirm + file. One issue, one prompt.
- Mode B (>3 sentences OR multi-item markers): decompose into candidates,
classify each as NEW / DUPLICATE_OF / ALREADY_RESOLVED / OUT_OF_SCOPE,
render approval table, file only the rows the operator approves. Never
silently drops a candidate.
- Mode C (ambiguous short input): exactly one clarifying question, no second.
Sensitivity sanitization (every body, both modes) per context-repo.md:
- public tier (most repos): strict strip — external individuals/orgs replaced
with role descriptors ("the funder", "a partner org rep"); internal URLs,
credentials, secrets removed; verbatim transcript/email content forbidden;
STAGE 13 confidentiality-leak self-check before emit.
- staff-ok tier (context repo): mild strip — internal staff names may stay,
external still stripped.
- private tier: no strip required, but never copy verbatim from a
confidential origin.
Hard rules (per pipeline-nevers.md):
- Never applies auto:approved, auto:auto-fixable, auto:requires-human, or
any override:* label — those are gates triage and humans own
- Never files a body containing raw transcript or email content
- Never silently drops a Mode B candidate
- Never infers repo from content keywords alone (explicit token / alias
whole-word match only — "there's an auth bug" does NOT become ai-security)
- --no-confirm is forbidden in Mode B (batch always requires approval)
- gh errors stop the run (no inline retry, no half-filed batches)
Bot-identity startup gate: refuses to run when gh auth status shows the
OpenGaryBot account active. The bot triages, the operator files. Mixing
creates a confusing audit trail per git-identity.md.
Repo aliases: bootstrap map for slingshot, platform, gary, no-animal-violence,
context, desloppify, graze, where-they-stand, ai-security, c4c-bootcamp,
mobius, avs, protein-research, ace, cryptpad-pm, n8n, privatemode-proxy,
api-gateway, project-compassionate-code, documentation, structured-coding,
.github. Operator extends over time.
Args: --repo overrides inference; --dry-run computes the full plan but never
files; --no-confirm skips the Mode A prompt (forbidden in Mode B).
Model: opus (mode detection, sensitivity reasoning, and candidate
extraction all benefit from frontier judgment; not mechanical enough for
Sonnet).
disable-model-invocation: true — operator-only.
Plus: extends .claude/rules/tooling-reference.md "Operator commands" section
with the /issue paragraph alongside /run /merge /unblock.
…zation, document run filename format, trim merge calibration cross-ref
OriginalGary
force-pushed
the
operator-cmd/issue
branch
from
b3c53f6 to
6b4739f
Compare
Contributor
Author
The startup gate manufactured an approval gate the pipeline doesn't need — `/issue` should run under whichever identity is active. Per auto-memory `feedback_issue_allows_bot_identity.md` and the new pipeline-nevers hard rule against manufactured gates, this section was re-introduced incorrectly. Drop it. Issue authorship under the bot is fine; the bot/operator split governs commits and automation only. Side effect: the only remaining unlabeled fence blocks were inside the deleted refusal text, so CodeRabbit's MD040 fence-language nits on this file are also resolved.
OriginalGary
dismissed stale reviews from coderabbitai[bot] and coderabbitai[bot]
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds
/issue, an operator-only slash command that takes anything from a one-sentence note to a pasted meeting transcript and turns it into well-formed GitHub issues thatscout/triagewould themselves accept. Auto-detects single-vs-batch — the operator never declares a mode.Same operator-command pattern as
/run//merge//unblock. Refuses to run under bot identity (the bot triages, the operator files).What it does
Mode A — input is ≤3 sentences AND a repo can be resolved → draft + sanitize + dedup-probe + confirm + file. One issue, one prompt.
Mode B — input is >3 sentences OR contains multi-item markers (bullets,
and also, paragraph breaks with topic shifts) → decompose into candidates, classify each as NEW / DUPLICATE_OF / ALREADY_RESOLVED / OUT_OF_SCOPE, render approval table, file only the rows the operator approves. Never silently drops a candidate —OUT_OF_SCOPErows show with a one-line reason.Mode C — ambiguous short input → exactly one clarifying question, no second.
Sensitivity sanitization (every body, both modes)
Per
~/.claude/rules/context-repo.md:the funder,a partner org rep); internal URLs, credentials, secrets removed; verbatim transcript/email content forbidden; STAGE 13 confidentiality-leak self-check before emit.Hard rules (per pipeline-nevers.md + context-repo.md)
auto:approved,auto:auto-fixable,auto:requires-human, or anyoverride:*label — those are gates triage and humans own; filing only setsstage:triagedplus type/severity/componentai-security. Misfiling is worse than asking--no-confirmis forbidden in Mode B (batch filing always requires explicit approval)gherrors stop the run (no inline retry, no half-filed batches; if the firstgh issue createin a batch fails, abort the rest and surface partial state explicitly)--force, no env-var override.Why bot-identity refusal
Issues filed by
/issuemust carry the operator's identity. Mixing operator-driven issue creation with bot identity creates a confusing audit trail (per~/.claude/rules/git-identity.mdthe bot is for commits/automation, not operator-driven issue creation). The bot triages; the operator files. Keep that line clean.Args
--repo <name>— explicit target repo. Overrides any inferred repo. In Mode B, applied as a default to all extracted candidates.--dry-run— compute the full plan (mode detection, repo inference, classification, draft) but never callgh issue create. Outputs everything that would be filed.--no-confirm— Mode A only: skip the confirm prompt and file immediately. Forbidden in Mode B.Repo aliases
Bootstrap map covering 24 nicknames (slingshot → slingshot-uk-phase1, platform, bot/gary, no-animal-violence, context, desloppify, graze, where-they-stand/wts, ai-security, c4c-bootcamp, mobius, avs, protein-research, ace, cryptpad-pm, n8n, privatemode-proxy, api-gateway, compassionate-code, documentation, structured-coding-with-ai, .github). Whole-word case-insensitive match. Operator extends over time.
Model
opus— mode detection, sensitivity reasoning, and candidate extraction all benefit from frontier judgment; not mechanical enough for Sonnet.disable-model-invocation: true— operator-only.Files changed
claude-code/.claude/skills/issue/SKILL.md— new skill (294 lines).claude/rules/tooling-reference.md— adds the ""Operator commands"" section with all four operator commands (/run, /merge, /unblock, /issue). Note: this section is also added by the open/run,/merge,/unblockPRs (feat(skills): /run — on-demand operator drive of the 14-stage pipeline #32, feat(skills): /merge — ranked merge queue with honest confidence scoring #33, feat(skills): /unblock — operator decisions queue, sensitivity-leak-safe #34); whichever merges first will require the others to rebase. This PR's version covers the full set.Test plan
The skill cannot be live-fired from the bot session that built it (the startup gate refuses, by design). Algorithm traces validated against the prompt's five test inputs:
'README links to a 404 on line 12' --repo .github→ drafts 3-section body, confirms, files. Title ≤80 chars; onlystage:triaged+type:docs+severity:trivial.--dry-run→ renders draft, appends(dry-run; nothing filed), nogh issue createcall.--no-confirmwould error out.'something's broken'→ one combined repo+intent question, no second question.'Dave from FoundationX said the dashboard is slow' --repo platform→ external individual + org name stripped, body uses ""a funder"" / ""the funder"".Closes nothing — there's no upstream issue for this work;
/issuewas specced directly.Summary by CodeRabbit
/run,/merge,/unblock, and/issuefor structured workflow control./issueskill for converting inputs into GitHub issues with mode-based drafting, approval workflows, repo inference, sanitization by sensitivity tier, and dry-run/confirmation safeguards.