@olegkoval/agent-skills 1.37.0 → 1.38.0

package/.github/copilot-instructions.md

@@ -33,6 +33,7 @@ Available reusable prompt files:
 - mvp-oneshot: .github/prompts/mvp-oneshot.prompt.md
 - starter-rules: .github/prompts/starter-rules.prompt.md
 - viral-launch: .github/prompts/viral-launch.prompt.md
+- context-repo: .github/prompts/context-repo.prompt.md
 - add-to-my-skills: .github/prompts/add-to-my-skills.prompt.md
 - promptctl: .github/prompts/promptctl.prompt.md
 - ai-tools-setup: .github/prompts/ai-tools-setup.prompt.md

package/.github/prompts/ai-tools-setup.prompt.md

@@ -6,7 +6,7 @@ description: "Set up, repair, and report on the RTK+ICM+Vox AI development toolk
 
 Use the olko:ai-tools-setup skill.
 
-# AI Tools Setup — RTK + ICM + Vox
+# AI Tools Setup: RTK + ICM + Vox
 
 Install, repair, and measure the RTK+ICM+Vox toolkit for Claude Code.
 
@@ -15,7 +15,7 @@ Install, repair, and measure the RTK+ICM+Vox toolkit for Claude Code.
 | Mode | When to use |
 |------|-------------|
 | **Setup** | First install on a new machine |
-| **Repair** | Something broke — hooks missing, MCP not loading |
+| **Repair** | Something broke: hooks missing, MCP not loading |
 | **Digest** | Weekly check: token savings, memory health, voice stats |
 
 ## Tools in the Bundle
@@ -24,7 +24,7 @@ Install, repair, and measure the RTK+ICM+Vox toolkit for Claude Code.
 |------|---------|---------|
 | **RTK** | latest | Compresses CLI output by ~89% before it reaches the model |
 | **ICM** | latest | Persistent SQLite memory across sessions (knowledge graph + hybrid search) |
-| **Vox** | latest | Spoken task notifications via local TTS — zero API calls |
+| **Vox** | latest | Spoken task notifications via local TTS (zero API calls) |
 
 ---
 
@@ -179,8 +179,8 @@ echo "=============================="
 
 ### What healthy output looks like
 
-- **RTK**: `Tokens saved: 5M+ (85%+)` across `500+ commands` — if savings are low, run `rtk discover` to find uncovered commands
-- **ICM**: Topics like `decisions-*`, `errors-resolved`, `preferences` with recent timestamps — if empty, start storing manually with `icm store`
+- **RTK**: `Tokens saved: 5M+ (85%+)` across `500+ commands`. If savings are low, run `rtk discover` to find uncovered commands
+- **ICM**: Topics like `decisions-*`, `errors-resolved`, `preferences` with recent timestamps. If empty, start storing manually with `icm store`
 - **Vox**: Non-zero speak count means the Stop hook fired
 
 ---
@@ -193,7 +193,7 @@ To get a weekly automated report without having to remember to run this skill:
 # Create a weekly digest script
 cat > ~/ai-tools-digest.sh << 'EOF'
 #!/bin/bash
-echo "=== Weekly AI Tools Digest — $(date) ===" | tee -a ~/ai-tools-digest.log
+echo "=== Weekly AI Tools Digest: $(date) ===" | tee -a ~/ai-tools-digest.log
 echo "" | tee -a ~/ai-tools-digest.log
 rtk gain 2>/dev/null | tee -a ~/ai-tools-digest.log
 echo "" | tee -a ~/ai-tools-digest.log
@@ -255,7 +255,7 @@ To unload: `launchctl unload ~/Library/LaunchAgents/com.oleg.ai-tools-digest.pli
 | `rtk gain` | Total tokens saved to date |
 | `rtk discover` | Find commands not yet covered by RTK |
 | `icm topics` | List all memory topic buckets |
-| `icm health` | Memory hygiene — stale, redundant, decay stats |
+| `icm health` | Memory hygiene: stale, redundant, decay stats |
 | `icm recall "query"` | Search memories semantically |
 | `icm store -t topic -c "content" -i high` | Store a memory manually |
 | `vox "text"` | Speak text immediately |

package/.github/prompts/apple-store-submit.prompt.md

@@ -12,7 +12,7 @@ Use the olko:apple-store-submit skill.
 
 ## Overview
 
-Parse Apple rejection feedback, implement targeted fixes, update App Store Connect metadata, and resubmit — covering the complete cycle from rejection email to approved build.
+Parse Apple rejection feedback, implement targeted fixes, update App Store Connect metadata, and resubmit, covering the complete cycle from rejection email to approved build.
 
 ## When to Use
 
@@ -28,9 +28,9 @@ Parse Apple rejection feedback, implement targeted fixes, update App Store Conne
 Read the full rejection email. Extract and list:
 
 - **Guideline numbers** violated (e.g., 2.1, 4.2.0, 5.1.1)
-- **Code-level issues** — missing entitlements, incorrect privacy strings, sandbox violations
-- **Metadata issues** — wrong screenshot sizes, subtitle policy, keyword stuffing
-- **UX issues** — crashes described by reviewer, broken flows, missing functionality
+- **Code-level issues**: missing entitlements, incorrect privacy strings, sandbox violations
+- **Metadata issues**: wrong screenshot sizes, subtitle policy, keyword stuffing
+- **UX issues**: crashes described by reviewer, broken flows, missing functionality
 
 Create a numbered fix plan: one item per rejection point, with the type of fix needed.
 
@@ -68,7 +68,7 @@ Create a numbered fix plan: one item per rejection point, with the type of fix n
 <true/>
 ```
 
-**Sandbox violations** — remove `NoEscapeBlock` patterns, replace deprecated APIs, use only sandboxed file access via Security-Scoped Bookmarks when needed.
+**Sandbox violations**: remove `NoEscapeBlock` patterns, replace deprecated APIs, use only sandboxed file access via Security-Scoped Bookmarks when needed.
 
 ### 4. Build and Archive
 
@@ -116,7 +116,7 @@ Common metadata fixes:
 In the submission form, fill **Notes for Reviewer**:
 
 - Explain each rejection point and what was changed
-- Reference specific guideline numbers: "Addressing 2.1 — added NSMicrophoneUsageDescription to Info.plist"
+- Reference specific guideline numbers: "Addressing 2.1: added NSMicrophoneUsageDescription to Info.plist"
 - If a feature requires special credentials to test, provide test account details
 
 ### 7. Submit and Track
@@ -125,7 +125,7 @@ In the submission form, fill **Notes for Reviewer**:
 2. In App Store Connect: select the new build, fill notes, **Submit for Review**
 3. Typical re-review: 24–48h
 4. If rejected again: respond in the **Resolution Center** thread (same case number)
-5. Escalate via Resolution Center for policy disagreements — do not open a new submission
+5. Escalate via Resolution Center for policy disagreements: do not open a new submission
 
 ## Build Number Protocol
 

package/.github/prompts/branch-cleanup.prompt.md

@@ -16,12 +16,12 @@ dry-run so you see exactly what will be deleted before anything is removed.
 ## Inputs
 
 - **Remote** (optional, default: `origin`)
-- **Base branch** (optional, default: `main` — detect from `gh repo view`)
+- **Base branch** (optional, default: `main`, detect from `gh repo view`)
 - **--dry-run** (optional, default: on): preview deletions without executing them
-- **--execute**: actually delete branches (requires explicit flag — safety default)
+- **--execute**: actually delete branches (requires explicit flag, safety default)
 - **--include-local**: also clean up merged local branches (off by default)
 
-## Step 1 — Fetch and prune remote tracking references
+## Step 1: Fetch and prune remote tracking references
 
 Always start here. This removes local tracking refs for remote branches that are
 already gone, without touching anything else:
@@ -37,7 +37,7 @@ After pruning, print the count of removed tracking refs:
 git branch -vv | grep ': gone]' | wc -l
 ```
 
-## Step 2 — Identify stale remote branches
+## Step 2: Identify stale remote branches
 
 Find remote branches whose associated PR is merged or closed:
 
@@ -53,7 +53,7 @@ git branch -r --merged "$REMOTE/$BASE" \
 ```
 
 Cross-reference with GitHub to confirm each branch's PR state. Only include branches
-with a confirmed merged or closed PR — an empty result means the branch is not eligible:
+with a confirmed merged or closed PR. An empty result means the branch is not eligible:
 
 ```bash
 STALE_BRANCHES=()
@@ -84,7 +84,7 @@ dependabot/npm_and_yarn/eslint-8.57.0    ←  43 merged 2026-07-31
 feature/login-refresh                    ←  38 merged 2026-07-28
 ```
 
-## Step 3 — Safety filter
+## Step 3: Safety filter
 
 **Never propose deletion of:**
 
@@ -103,14 +103,14 @@ OPEN=$(gh pr list --head "$branch" --state open --json number --jq 'length')
 [ "$OPEN" -gt 0 ] && echo "SKIP (open PR): $branch"
 ```
 
-## Step 4 — Preview or execute
+## Step 4: Preview or execute
 
 ### Dry-run (default)
 
 Print the deletion plan and stop. Never touch the remote without `--execute`:
 
 ```
-Branch cleanup preview — dry-run (pass --execute to apply)
+Branch cleanup preview, dry-run (pass --execute to apply)
 
   Remote branches to delete (all merged to main):
     origin/dependabot/npm_and_yarn/lodash-4.17.21     (PR #42, merged 2026-07-31)
@@ -135,7 +135,7 @@ for branch in "${STALE_BRANCHES[@]}"; do
 done
 ```
 
-## Step 5 — Local branch cleanup (optional, --include-local)
+## Step 5: Local branch cleanup (optional, --include-local)
 
 Only runs when `--include-local` is explicitly passed.
 
@@ -153,10 +153,10 @@ gone before deleting:
 
 ```bash
 git branch -vv "$branch"   # should show ': gone]' if remote is deleted
-git branch -d "$branch"    # safe delete — fails if unmerged work exists
+git branch -d "$branch"    # safe delete: fails if unmerged work exists
 ```
 
-Never use `-D` (force delete) — let `-d` protect you from accidental data loss.
+Never use `-D` (force delete): let `-d` protect you from accidental data loss.
 
 ## Report
 
@@ -164,7 +164,7 @@ Never use `-D` (force delete) — let `-d` protect you from accidental data loss
 Branch cleanup complete
 
   Remote:
-    Deleted:   12 branches (10 Dependabot, 2 feature branches — all merged)
+    Deleted:   12 branches (10 Dependabot, 2 feature branches, all merged)
     Protected: 2 skipped (beta, open PR)
 
   Local tracking refs:
@@ -183,4 +183,4 @@ Branch cleanup complete
 |---|---|
 | `olko:dependabot-triage` (merged patch PRs) | next `olko:morning-routine` starts clean |
 | Sprint wind-down / milestone close | `git fetch` confirms clean state |
-| Any large batch merge | nothing — maintenance complete |
+| Any large batch merge | nothing: maintenance complete |

package/.github/prompts/ci-fix-loop.prompt.md

@@ -35,7 +35,7 @@ gh pr checks --json name,state,conclusion 2>/dev/null \
 
 Classify each check as: `pass`, `fail`, `pending`, or `skipped`.
 
-**Exit immediately if all required checks already pass** — nothing to do.
+**Exit immediately if all required checks already pass**: nothing to do.
 
 ### 2. Loop (max 5 iterations)
 
@@ -48,7 +48,7 @@ gh pr checks --json name,state,conclusion \
 
 For each failing check, record: check name, workflow file, job name.
 
-Stop if the working set is empty — all required checks are green.
+Stop if the working set is empty: all required checks are green.
 
 #### B. Check for base-branch regression first
 
@@ -59,7 +59,7 @@ gh run list --branch <base-branch> --workflow <workflow-name> --limit 3 \
   --json conclusion --jq '.[].conclusion'
 ```
 
-If the same check is failing on the base branch, **stop and report** —
+If the same check is failing on the base branch, **stop and report**:
 this is not your PR's fault. Do not attempt a fix that masks a base-branch problem.
 Wait for a `Base branch recovered` notice before re-trying.
 
@@ -76,7 +76,7 @@ gh run list --branch <branch> --workflow <workflow-name> --limit 1 \
 gh run view <run-id> --log-failed
 ```
 
-Cap log reading at ~200 lines per job — the error is almost always in the last
+Cap log reading at ~200 lines per job: the error is almost always in the last
 50 lines of a failed step.
 
 #### D. Diagnose
@@ -128,12 +128,12 @@ go mod tidy          # Go
 ```
 
 For test failures: read the failing test and the code it exercises. Fix the code
-or the assertion — but not both in one commit unless the test was clearly wrong
+or the assertion, but not both in one commit unless the test was clearly wrong
 and the fix is trivial.
 
 #### F. Commit and push
 
-Stage **only the files touched in step E** — never `git add -A`, which can sweep
+Stage **only the files touched in step E**: never `git add -A`, which can sweep
 in unrelated local changes:
 
 ```bash
@@ -171,7 +171,7 @@ Go back to step A.
 ```
 CI fix loop complete.
   Iterations:  2
-  Fixed:       lint (gofmt: 3 files), test (TestUserCreate — nil pointer in fixture)
+  Fixed:       lint (gofmt: 3 files), test (TestUserCreate: nil pointer in fixture)
   Blocked:     none
   Status:      all required checks green
 ```

package/.github/prompts/cloudflare-block-countries.prompt.md

@@ -37,7 +37,7 @@ curl -s -X GET \
   jq '.result[] | select(.phase == "http_request_firewall_custom")'
 ```
 
-If the output is empty — no existing ruleset. Go to **3a**. If a ruleset exists, note its `id` and go to **3b**.
+If the output is empty: no existing ruleset. Go to **3a**. If a ruleset exists, note its `id` and go to **3b**.
 
 ### 3a. Create NEW Ruleset with Block Rule
 

package/.github/prompts/coderabbitloop.prompt.md

@@ -1,7 +1,7 @@
 <!-- Generated by scripts/build-adapters.sh. Do not edit directly. -->
 
 ---
-description: "Iteratively drives a GitHub PR to zero unresolved CodeRabbit findings — reads each inline comment's own Prompt for AI Agents block, applies the fix, replies to the thread, resolves it, pushes, and re-triggers CodeRabbit."
+description: "Iteratively drives a GitHub PR to zero unresolved CodeRabbit findings: reads each inline comment's own Prompt for AI Agents block, applies the fix, replies to the thread, resolves it, pushes, and re-triggers CodeRabbit."
 ---
 
 Use the olko:coderabbitloop skill.
@@ -14,21 +14,21 @@ Fix every actionable CodeRabbit finding on a PR, answer each thread, resolve it,
 
 - **PR number** (optional): detect from the current branch if not given.
 
-## How CodeRabbit actually posts (verified against docs.coderabbit.ai and a live install — don't assume otherwise)
+## How CodeRabbit actually posts (verified against docs.coderabbit.ai and a live install: don't assume otherwise)
 
 CodeRabbit runs as `coderabbitai[bot]` and posts **two independent things**, not one:
 
-1. An auto-generated **walkthrough/summary issue comment** — body starts with the HTML marker `<!-- This is an auto-generated comment: summarize by coderabbit.ai -->`. This is a description, not a review — use it for context only, never as a source of fixable findings.
+1. An auto-generated **walkthrough/summary issue comment**: body starts with the HTML marker `<!-- This is an auto-generated comment: summarize by coderabbit.ai -->`. This is a description, not a review: use it for context only, never as a source of fixable findings.
 2. A **formal PR review** carrying one **inline review comment per finding**, each a real, resolvable GitHub review thread at the finding's file/line. Each one contains a `<details><summary>🤖 Prompt for AI Agents</summary>` block: a ready-made fix spec, CodeRabbit's own words, meant to be handed straight to an agent.
 
-**Use the inline threads as the source of truth**, exactly like Greptile/Qodo — fetch via GraphQL `reviewThreads`, which carries native `isResolved` state (see `references/graphql-queries.md`, shared with `qodoloop`; only the bot login and the prompt-heading text differ).
+**Use the inline threads as the source of truth**, exactly like Greptile/Qodo: fetch via GraphQL `reviewThreads`, which carries native `isResolved` state (see `references/graphql-queries.md`, shared with `qodoloop`; only the bot login and the prompt-heading text differ).
 
-CodeRabbit has slash commands (official, `docs.coderabbit.ai/reference/review-commands`) — the ones this loop actually uses:
-- `@coderabbitai review` — incremental review of new changes (use this to re-trigger after a push, if automatic review-on-push is off for this repo).
-- `@coderabbitai resolve` — marks **all** CodeRabbit threads resolved in one shot. This is a bulk escape hatch, not the default path: this loop resolves threads one at a time, right after replying to each, so the reply-then-resolve record stays intact per finding.
-- CodeRabbit also has its own built-in `@coderabbitai autofix`, which commits fixes directly without a model in the loop. Mention it to the user as an alternative if they'd rather skip the loop entirely, but this skill's whole point is the Claude-orchestrated version — don't default to autofix instead of doing the work.
+CodeRabbit has slash commands (official, `docs.coderabbit.ai/reference/review-commands`), the ones this loop actually uses:
+- `@coderabbitai review`: incremental review of new changes (use this to re-trigger after a push, if automatic review-on-push is off for this repo).
+- `@coderabbitai resolve`: marks **all** CodeRabbit threads resolved in one shot. This is a bulk escape hatch, not the default path: this loop resolves threads one at a time, right after replying to each, so the reply-then-resolve record stays intact per finding.
+- CodeRabbit also has its own built-in `@coderabbitai autofix`, which commits fixes directly without a model in the loop. Mention it to the user as an alternative if they'd rather skip the loop entirely, but this skill's whole point is the Claude-orchestrated version: don't default to autofix instead of doing the work.
 
-**Rate limit is a real terminal state, not a bug.** A comment reading "you've reached your PR review limit" (confirmed live wording) means CodeRabbit will not review again this cycle — stop and report it, don't loop waiting for something that isn't coming.
+**Rate limit is a real terminal state, not a bug.** A comment reading "you've reached your PR review limit" (confirmed live wording) means CodeRabbit will not review again this cycle: stop and report it, don't loop waiting for something that isn't coming.
 
 ## Instructions
 
@@ -54,11 +54,11 @@ If this repo doesn't auto-review on push, nudge it:
 gh pr comment <PR_NUMBER> --body "@coderabbitai review"
 ```
 
-Poll (every ~10s, timeout ~5min) for CodeRabbit's next formal review to land — a new entry in `gh api repos/{owner}/{repo}/pulls/<PR_NUMBER>/reviews` from `coderabbitai[bot]` created after this push. If you instead see a rate-limit comment ("review limit"), **stop the loop now** and go to Report with `blocked: rate-limited`.
+Poll (every ~10s, timeout ~5min) for CodeRabbit's next formal review to land: a new entry in `gh api repos/{owner}/{repo}/pulls/<PR_NUMBER>/reviews` from `coderabbitai[bot]` created after this push. If you instead see a rate-limit comment ("review limit"), **stop the loop now** and go to Report with `blocked: rate-limited`.
 
 #### B. Fetch unresolved findings
 
-Run the `unresolvedQodoThreads`-shaped query in `references/graphql-queries.md` (filter `author.login == "coderabbitai[bot]"` instead of Qodo's login). For each unresolved thread, parse the finding text and the **Prompt for AI Agents** block (heading text is `🤖 Prompt for AI Agents`, not `Agent Prompt` — match case/emoji-insensitively).
+Run the `unresolvedQodoThreads`-shaped query in `references/graphql-queries.md` (filter `author.login == "coderabbitai[bot]"` instead of Qodo's login). For each unresolved thread, parse the finding text and the **Prompt for AI Agents** block (heading text is `🤖 Prompt for AI Agents`, not `Agent Prompt`: match case/emoji-insensitively).
 
 #### C. Check exit conditions
 
@@ -66,17 +66,17 @@ Stop if there are zero unresolved CodeRabbit threads, or max iterations reached.
 
 #### D. Fix each finding
 
-For each thread, in order: read the Prompt for AI Agents block in full — it names the issue and the fix. Apply it, staying inside its stated scope. If the suggestion is wrong, unsafe, or needs a product decision, don't force it — note that in step E instead.
+For each thread, in order: read the Prompt for AI Agents block in full: it names the issue and the fix. Apply it, staying inside its stated scope. If the suggestion is wrong, unsafe, or needs a product decision, don't force it: note that in step E instead.
 
 #### E. Answer and resolve
 
-**Reply to the thread first, then resolve it** — the "answer to the comments" half of the job:
+**Reply to the thread first, then resolve it**, the "answer to the comments" half of the job:
 
 ```bash
 gh api graphql -f query='mutation { addPullRequestReviewThreadReply(input: {pullRequestReviewThreadId: "<THREAD_ID>", body: "<REPLY>"}) { comment { id } } }'
 ```
 
-Reply body: one or two sentences — what you changed and why, or exactly why you didn't fix it. Then resolve (`resolveReviewThread`). A finding you deliberately skip still gets a reply and gets resolved.
+Reply body: one or two sentences, what you changed and why, or exactly why you didn't fix it. Then resolve (`resolveReviewThread`). A finding you deliberately skip still gets a reply and gets resolved.
 
 #### F. Commit and push
 
@@ -107,4 +107,4 @@ Coderabbitloop complete.
 
 ## References
 
-- `references/graphql-queries.md` — shared with `qodoloop`: the exact `reviewThreads` query and the `resolveReviewThread` / `addPullRequestReviewThreadReply` mutations. Swap the author-login filter to `coderabbitai[bot]` and the prompt-heading match to `Prompt for AI Agents`.
+- `references/graphql-queries.md`: shared with `qodoloop`: the exact `reviewThreads` query and the `resolveReviewThread` / `addPullRequestReviewThreadReply` mutations. Swap the author-login filter to `coderabbitai[bot]` and the prompt-heading match to `Prompt for AI Agents`.

package/.github/prompts/codexloop.prompt.md

@@ -1,14 +1,14 @@
 <!-- Generated by scripts/build-adapters.sh. Do not edit directly. -->
 
 ---
-description: "Iteratively drives a GitHub PR to zero unresolved OpenAI Codex review comments — but verifies every finding against the real code first, fixes only the correct ones, and rebuts false positives without changing correct code."
+description: "Iteratively drives a GitHub PR to zero unresolved OpenAI Codex review comments, verifying every finding against the real code first, fixes only the correct ones, and rebuts false positives without changing correct code."
 ---
 
 Use the olko:codexloop skill.
 
 # Codexloop
 
-Drive a GitHub PR until Codex has no unresolved review comments — **but do not cargo-cult its
+Drive a GitHub PR until Codex has no unresolved review comments, **but do not cargo-cult its
 suggestions.** Every comment is a *claim to verify*, not an instruction to obey. A wrong
 suggestion applied is worse than the comment itself.
 
@@ -47,10 +47,10 @@ Pick the login matching `*codex*` (commonly `chatgpt-codex-connector[bot]`, some
 `codex[bot]`) and export it as `BOT`.
 
 **Finding nothing here is NOT a stop condition.** On a PR Codex has never reviewed there is no bot
-login to find yet — that is the normal starting state, not a missing app. When the probe comes back
+login to find yet: that is the normal starting state, not a missing app. When the probe comes back
 empty, fall through to step 2A, post `@codex review`, and re-run this probe once a review lands;
 until then match any login containing `codex` when polling. Only conclude the app is absent after
-step 2A's bounded wait has expired with no review and no codex-like login anywhere on the PR — and
+step 2A's bounded wait has expired with no review and no codex-like login anywhere on the PR, and
 then say so and stop, rather than looping against nothing.
 
 ## 1. Identify the PR
@@ -63,7 +63,7 @@ Switch to the PR branch if not already on it. Capture `OWNER`/`REPO` (`gh repo v
 
 ## 2. The loop (max 5 iterations)
 
-Keep an explicit iteration counter and stop at 5 — the cap is a real bound to enforce, not a
+Keep an explicit iteration counter and stop at 5: the cap is a real bound to enforce, not a
 figure of speech. Each pass through A–G is one iteration; on hitting the cap, go straight to the
 report and list what is still unresolved rather than starting a sixth.
 
@@ -77,13 +77,13 @@ HAVE=$(gh api repos/{owner}/{repo}/pulls/<PR>/reviews --paginate \
 if [ "$HAVE" = "0" ]; then gh pr comment <PR> --body "@codex review"; fi
 ```
 
-Poll for the review of THIS head to land. No check-run exists, so poll the reviews endpoint — and
+Poll for the review of THIS head to land. No check-run exists, so poll the reviews endpoint, and
 poll it on a **deadline**, never `while true`: a review that never arrives must end the skill with
 an honest timeout, not hang it.
 
 ```bash
 # 10-minute deadline, one retry, then give up. DEADLINE/RETRIES are the
-# enforcement of the bounds this skill claims — do not drop them.
+# enforcement of the bounds this skill claims: do not drop them.
 wait_for_review() {                       # $1 = attempt label
   local deadline=$(( SECONDS + 600 ))
   while [ "$SECONDS" -lt "$deadline" ]; do
@@ -96,7 +96,7 @@ wait_for_review() {                       # $1 = attempt label
 }
 
 if ! wait_for_review "first wait"; then
-  echo "no Codex review after 10m — retrying once"     # say the retry out loud
+  echo "no Codex review after 10m, retrying once"     # say the retry out loud
   gh pr comment <PR> --body "@codex review"
   if ! wait_for_review "after retry"; then
     echo "Codex did not review $HEAD_SHA after a retry; stopping and reporting."
@@ -110,7 +110,7 @@ retry in the final summary; two silent timeouts are the failure mode this guard
 
 ### B. Fetch the findings
 
-- **Summary**: the review `.body` from the object above — read the overall take and the priority
+- **Summary**: the review `.body` from the object above: read the overall take and the priority
   spread.
 - **Unresolved inline comments** on the current head:
 
@@ -125,11 +125,11 @@ unresolved ones.
 ### C. Critically evaluate EACH comment (the core of this skill)
 
 For every comment, **verify the claim against the actual code and repo conventions before touching
-anything.** Read the whole file — not just the diff hunk Codex saw — plus the types and the call
+anything.** Read the whole file, not just the diff hunk Codex saw, plus the types and the call
 sites. Then classify:
 
-1. **CORRECT + actionable** — the finding is real and the fix improves the code. → fix it (step D).
-2. **FALSE POSITIVE / technically wrong** — the claim doesn't hold. → do **NOT** change code; write a
+1. **CORRECT + actionable**: the finding is real and the fix improves the code. → fix it (step D).
+2. **FALSE POSITIVE / technically wrong**: the claim doesn't hold. → do **NOT** change code; write a
    specific, evidence-based reply (cite the exact code/line/behavior that disproves it), then resolve.
 3. **Valid but out-of-scope / stylistic nit** that conflicts with repo convention or the PR's intent
    → briefly decline with a reason, then resolve. Do not expand the PR's scope to satisfy a nit.
@@ -152,7 +152,7 @@ sites. Then classify:
 - Restating library/framework semantics incorrectly.
 - Style demands that contradict the repo's existing, consistent pattern.
 
-### D. Apply fixes — category 1 only
+### D. Apply fixes: category 1 only
 
 Make the minimal correct change. Re-run the local gate if the repo has one (typecheck/tests) before
 moving on.
@@ -160,13 +160,13 @@ moving on.
 ### E. Commit and push FIRST, before resolving anything
 
 Order matters. A resolved thread is a claim that the fix is on the branch, so the push has to
-succeed before the claim is made — otherwise a failed commit or push leaves the PR unfixed with the
+succeed before the claim is made: otherwise a failed commit or push leaves the PR unfixed with the
 finding marked resolved, and nobody looks at it again.
 
 If step D changed code:
 
 ```bash
-# Stage ONLY the files your fixes touched — never `git add -A`, which sweeps up
+# Stage ONLY the files your fixes touched: never `git add -A`, which sweeps up
 # unrelated work and untracked secrets sitting in the worktree.
 git status --short                     # look before you stage
 git add <path> [<path>...]             # the files named in the findings you fixed
@@ -187,7 +187,7 @@ If they differ, stop: the fix is not on the PR, so nothing may be resolved yet.
 ### F. Reply to and resolve every addressed thread
 
 Only now, with the fixes pushed, reply and resolve. Fetch unresolved threads, **following
-pagination** — a PR with more than 100 threads will otherwise look clean while unresolved findings
+pagination**: a PR with more than 100 threads will otherwise look clean while unresolved findings
 sit on page two:
 
 ```bash
@@ -219,7 +219,7 @@ then resolve:
 gh api graphql -f query='mutation { resolveReviewThread(input: {threadId: "THREAD_ID"}) { thread { isResolved } } }'
 ```
 
-Resolve a thread only for comments authored by `$BOT` that you have fixed or rebutted — never
+Resolve a thread only for comments authored by `$BOT` that you have fixed or rebutted: never
 blanket-resolve, and never resolve a human reviewer's thread.
 
 Threads you are **rebutting** need no push, so they may be replied to and resolved regardless of
@@ -235,9 +235,9 @@ ensure all threads are resolved, and exit.
 
 Stop when **any** is true:
 - Zero unresolved `$BOT` comments remain, and every comment this round was fixed or
-  rebutted+resolved. (There is no score to hit — this is "done".)
-- Max iterations (5) reached — report what remains.
-- Codex never responded after one retry — report the timeout honestly; do not claim success.
+  rebutted+resolved. (There is no score to hit: this is "done".)
+- Max iterations (5) reached: report what remains.
+- Codex never responded after one retry: report the timeout honestly; do not claim success.
 
 ## 4. Report