@willbooster/agent-skills 1.23.6 → 1.24.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (6) hide show
  1. package/dist/cli.js +3 -3
  2. package/package.json +1 -1
  3. package/skills/review-fix-all/SKILL.md +15 -6
  4. package/skills/review-fix-claude/SKILL.md +16 -5
  5. package/skills/review-fix-codex/SKILL.md +16 -5
  6. package/skills/review-fix-gemini/SKILL.md +16 -5
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@willbooster/agent-skills",
3
- "version": "1.23.6",
3
+ "version": "1.24.0",
4
4
  "description": "A collection of agent skills for WillBooster Inc.",
5
5
  "repository": {
6
6
  "type": "git",
package/skills/review-fix-all/SKILL.md CHANGED
@@ -7,11 +7,20 @@ allowed-tools: Bash(bunx:*), Task
7
7
  # Review workflow
8
8
 
9
9
  1. Run the following command with a 1-hour timeout (RUN IT AS A BLOCKING, FOREGROUND TASK TO PREVENT PREMATURE TERMINATION): `bunx @willbooster/agent-skills@latest review --agent all`
10
- 2. Treat the combined output as candidate review comments generated concurrently by multiple agents.
11
- 3. Merge the review results into a single set of candidate comments, deduplicating those that address the same underlying issue.
12
- 4. If there are no candidate comments, quit the workflow without modifying the code.
13
- 5. Determine whether each candidate comment is still valid for the current codebase.
14
- 6. Address valid review comments by updating the code. For invalid comments, add explanatory comments to the code detailing why the existing implementation is necessary.
10
+ 2. Merge the returned findings, deduplicating those that address the same underlying issue. If there are no findings, quit without modifying the code.
11
+ 3. Accumulate every returned finding as a fact inside `## Verified Facts`, together with any returned `## Verified Facts` sections. Keep only facts that still affect review: what is known, why it matters, and how it was verified.
12
+ 4. Refine accumulated facts to reduce duplication and remove stale details, then pass them through stdin with heredoc syntax:
13
+
14
+ ```bash
15
+ bunx @willbooster/agent-skills@latest review --agent all <<'EOF'
16
+ ## Verified Facts
17
+
18
+ [all accumulated facts]
19
+ EOF
20
+ ```
21
+
22
+ 5. Determine whether each finding is still valid for the current codebase, reusing accumulated facts.
23
+ 6. Address valid findings by updating the code. For invalid findings, add explanatory comments to the code detailing why the existing implementation is necessary.
15
24
  - If there are valid review comments that are out of scope for the current PR, create new issues for them.
16
25
  7. If you made any code changes, commit and push them to the current branch.
17
- 8. If you fixed any valid comments, return to step 1. Otherwise, quit the workflow.
26
+ 8. If you fixed any valid findings, return to step 4. Otherwise, quit the workflow.
package/skills/review-fix-claude/SKILL.md CHANGED
@@ -7,9 +7,20 @@ allowed-tools: Bash(bunx:*)
7
7
  # Review workflow
8
8
 
9
9
  1. Run the following command with a 1-hour timeout (RUN IT AS A BLOCKING, FOREGROUND TASK TO PREVENT PREMATURE TERMINATION): `bunx @willbooster/agent-skills@latest review --agent claude`
10
- 2. Treat the returned review results as the candidate comment set to process. If the command returns `There is no concern.`, quit without modifying code.
11
- 3. Determine whether each candidate comment is still valid for the current codebase.
12
- 4. Address valid review comments by updating the code. For invalid comments, add explanatory comments to the code detailing why the existing implementation is necessary.
10
+ 2. If the command returns `There is no concern.`, quit without modifying code. Otherwise, process the returned findings.
11
+ 3. Accumulate every returned finding as a fact inside `## Verified Facts`, together with any returned `## Verified Facts`. Keep only facts that still affect review: what is known, why it matters, and how it was verified.
12
+ 4. Refine accumulated facts to reduce duplication and remove stale details, then pass them through stdin with heredoc syntax:
13
+
14
+ ```bash
15
+ bunx @willbooster/agent-skills@latest review --agent claude <<'EOF'
16
+ ## Verified Facts
17
+
18
+ [all accumulated facts]
19
+ EOF
20
+ ```
21
+
22
+ 5. Determine whether each finding is still valid for the current codebase, reusing accumulated facts.
23
+ 6. Address valid findings by updating the code. For invalid findings, add explanatory comments to the code detailing why the existing implementation is necessary.
13
24
  - If there are valid review comments that are out of scope for the current PR, create new issues for them.
14
- 5. If you made any code changes, commit and push them to the current branch.
15
- 6. If you fixed any valid comments, return to step 1. Otherwise, quit the workflow.
25
+ 7. If you made any code changes, commit and push them to the current branch.
26
+ 8. If you fixed any valid findings, return to step 4. Otherwise, quit the workflow.
package/skills/review-fix-codex/SKILL.md CHANGED
@@ -7,9 +7,20 @@ allowed-tools: Bash(bunx:*)
7
7
  # Review workflow
8
8
 
9
9
  1. Run the following command with a 1-hour timeout (RUN IT AS A BLOCKING, FOREGROUND TASK TO PREVENT PREMATURE TERMINATION): `bunx @willbooster/agent-skills@latest review --agent codex`
10
- 2. Treat the returned review results as the candidate comment set to process. If the command returns `There is no concern.`, quit without modifying code.
11
- 3. Determine whether each candidate comment is still valid for the current codebase.
12
- 4. Address valid review comments by updating the code. For invalid comments, add explanatory comments to the code detailing why the existing implementation is necessary.
10
+ 2. If the command returns `There is no concern.`, quit without modifying code. Otherwise, process the returned findings.
11
+ 3. Accumulate every returned finding as a fact inside `## Verified Facts`, together with any returned `## Verified Facts`. Keep only facts that still affect review: what is known, why it matters, and how it was verified.
12
+ 4. Refine accumulated facts to reduce duplication and remove stale details, then pass them through stdin with heredoc syntax:
13
+
14
+ ```bash
15
+ bunx @willbooster/agent-skills@latest review --agent codex <<'EOF'
16
+ ## Verified Facts
17
+
18
+ [all accumulated facts]
19
+ EOF
20
+ ```
21
+
22
+ 5. Determine whether each finding is still valid for the current codebase, reusing accumulated facts.
23
+ 6. Address valid findings by updating the code. For invalid findings, add explanatory comments to the code detailing why the existing implementation is necessary.
13
24
  - If there are valid review comments that are out of scope for the current PR, create new issues for them.
14
- 5. If you made any code changes, commit and push them to the current branch.
15
- 6. If you fixed any valid comments, return to step 1. Otherwise, quit the workflow.
25
+ 7. If you made any code changes, commit and push them to the current branch.
26
+ 8. If you fixed any valid findings, return to step 4. Otherwise, quit the workflow.
package/skills/review-fix-gemini/SKILL.md CHANGED
@@ -7,9 +7,20 @@ allowed-tools: Bash(bunx:*)
7
7
  # Review workflow
8
8
 
9
9
  1. Run the following command with a 1-hour timeout (RUN IT AS A BLOCKING, FOREGROUND TASK TO PREVENT PREMATURE TERMINATION): `bunx @willbooster/agent-skills@latest review --agent gemini`
10
- 2. Treat the returned review results as the candidate comment set to process. If the command returns `There is no concern.`, quit without modifying code.
11
- 3. Determine whether each candidate comment is still valid for the current codebase.
12
- 4. Address valid review comments by updating the code. For invalid comments, add explanatory comments to the code detailing why the existing implementation is necessary.
10
+ 2. If the command returns `There is no concern.`, quit without modifying code. Otherwise, process the returned findings.
11
+ 3. Accumulate every returned finding as a fact inside `## Verified Facts`, together with any returned `## Verified Facts`. Keep only facts that still affect review: what is known, why it matters, and how it was verified.
12
+ 4. Refine accumulated facts to reduce duplication and remove stale details, then pass them through stdin with heredoc syntax:
13
+
14
+ ```bash
15
+ bunx @willbooster/agent-skills@latest review --agent gemini <<'EOF'
16
+ ## Verified Facts
17
+
18
+ [all accumulated facts]
19
+ EOF
20
+ ```
21
+
22
+ 5. Determine whether each finding is still valid for the current codebase, reusing accumulated facts.
23
+ 6. Address valid findings by updating the code. For invalid findings, add explanatory comments to the code detailing why the existing implementation is necessary.
13
24
  - If there are valid review comments that are out of scope for the current PR, create new issues for them.
14
- 5. If you made any code changes, commit and push them to the current branch.
15
- 6. If you fixed any valid comments, return to step 1. Otherwise, quit the workflow.
25
+ 7. If you made any code changes, commit and push them to the current branch.
26
+ 8. If you fixed any valid findings, return to step 4. Otherwise, quit the workflow.