Brainstorm Agent Runs

Skill: brainstorm-agent-runs

Analyze execution logs of agentic runs in CI/CD pipelines to extract insights about implementation status, challenges, and next steps for a Pull Request.

When to Use

• When analyzing a Pull Request and you need to understand the outcome of associated agentic workflow runs.
• If an agent task was triggered on a PR but the results are unclear or hidden within CI check details.
• To map out the execution history and failures of agents interacting with a specific Pull Request.

When Not to Use

• For analyzing non-agent CI pipelines (e.g., standard unit tests or linters), where github-actions or gh-run skills are more direct.
• When generating diagrams for codebase architecture or generic GitHub PR history (use brainstorm-github-pr instead).
• If the repository does not employ GitHub Agentic Workflows (gh-aw).

Common Pitfalls

• Using gh pr checks Exclusively: Relying solely on gh pr checks and missing workflow_dispatch or issue_comment triggers that aren't natively attached to the HEAD commit.
• Fetching Full Logs Too Early: Trying to download the complete execution logs for every run before creating the high-level Ishikawa diagram, wasting token budget.
• Ignoring Branch Context: Querying the /actions/runs API without filtering by the PR's headRefName or title, returning unrelated runs from main or other branches.

Core Process

1. Trigger Recognition: Activate when an active Pull Request is associated with the runtime context or the user requests PR analysis.
2. Identify Runs: Use the GitHub API to query all runs matching either the PR branch OR title, bypassing gh pr checks limitations.
3. Extract Insights: Analyze the identified runs (successes, failures, missing facts) without checking detailed logs prematurely.
4. Visualize: Generate an ishikawa-beta diagram representing the findings.

Commands / Usage Patterns

List all agent runs for a PR by matching either branch name or PR title:

branch_name=$(gh pr view  --repo / --json headRefName -q .headRefName)
pr_title=$(gh pr view  --repo / --json title -q .title)

gh api repos///actions/runs --paginate \
  -q ".workflow_runs[] | select((.head_branch == \"$branch_name\" or .display_title == \"$pr_title\") and .name == \"Cogni AI Agent\") | {id: .id, status: .status, conclusion: .conclusion, event: .event}"

Brainstorming - Agent runs

When an active Pull Request is associated with the runtime context or the user requests PR analysis, you MUST identify agentic runs in the CI/CD pipeline and analyze their execution logs to extract insights about the implementation status, challenges, and next steps.

Step 1: Agent PR runs

First, identify any agentic runs in the CI/CD pipeline associated with the Pull Request. Because gh pr checks only shows the HEAD commit's latest runs and routinely misses manual workflow calls (workflow_dispatch) or keyword triggers (issue_comment), you MUST use the GitHub API to query all runs matching either the PR branch OR title.

Command to list all agent runs for a PR:

branch_name=$(gh pr view  --repo / --json headRefName -q .headRefName)
pr_title=$(gh pr view  --repo / --json title -q .title)

gh api repos///actions/runs --paginate \
  -q ".workflow_runs[] \
  | select((.head_branch == \"$branch_name\" or .display_title == \"$pr_title\") and .name == \"Cogni AI Agent\") \
  | {id: .id, status: .status, conclusion: .conclusion, event: .event}"

Example ishikawa-beta Diagram:

%% This diagram visualizes the identification of agentic runs in the CI/CD pipeline for a PR.
ishikawa-beta
    Cogni AI Agent Runs
    issue_comment
        Agent Run (#)
            success
    workflow_dispatch
        Agent Run (#)
            failure
            Missing facts

At this step, don't check for more detailed logs yet.

Related Skills

• brainstorm:

You MUST load this skill when asked to brainstorm, explore options, or break down complex problems.

• brainstorm-github-pr:

You MUST load this skill when asked to analyze or brainstorm a Pull Request.

• gh-api:

You MUST load this skill when executing advanced GitHub CLI API queries.

• mermaid:

You MUST load this skill when creating or maintaining stable Mermaid.js diagrams.

• mermaid-beta:

You MUST load this skill when creating or maintaining experimental Mermaid.js beta diagrams like ishikawa-beta.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Cogni-AI-OU
• Source: Cogni-AI-OU/cogni-ai-agent-skills
• License: MIT

Install and usage instructions live in the source repository linked above.