ci: fix 'Argument list too long' in release-on-main workflow

[benvinegar]

Problem

The release workflow fails with:

An error occurred trying to start process '/usr/bin/bash' with working directory
'/home/runner/work/baudbot/baudbot'. Argument list too long

Root cause: PR context (140+ merged PRs with full bodies = ~155KB of JSON) was passed through GITHUB_OUTPUT and then injected as PR_CONTEXT env vars in the "Decide release bump" and "Build changelog" steps. This exceeded Linux ARG_MAX, preventing bash from starting at all.

Since there are no release tags yet, the workflow gathered every merged PR since epoch.

Fix

1. Write PR context to a file (/tmp/pr_context.json) instead of passing through GITHUB_OUTPUT → env vars
2. Read from file in downstream steps (jq --slurpfile for the Anthropic payload, direct path for changelog)
3. Truncate PR bodies to 200 chars — Claude only needs titles and a brief summary, not full PR descriptions
4. Remove PR_CONTEXT env vars from both the decide and changelog steps

Verification

• ✅ yaml-lint passes
• ✅ actionlint passes (only shellcheck style suggestions, no errors)
• ✅ jq --slurpfile behavior verified locally
• ✅ Body truncation + null-body handling verified locally

To verify on CI without merging

You can test this workflow on the branch by triggering it manually:

gh workflow run release-on-main.yml --ref fix/release-workflow-arg-list-too-long

(Requires workflow_dispatch trigger, which is already configured)

[greptile-apps]

Greptile Summary

Fixes Linux ARG_MAX error by writing PR context to /tmp/pr_context.json instead of passing through GITHUB_OUTPUT environment variables. PR bodies are truncated to 200 characters to reduce payload size. The jq --slurpfile approach correctly loads the file and accesses it via $prs[0] in the Anthropic API payload construction.

• Removes pr_context output from GITHUB_OUTPUT and writes directly to file (line 81, 86)
• Removes PR_CONTEXT env vars from both decide and changelog steps
• Uses jq --slurpfile prs /tmp/pr_context.json with $prs[0] array access (line 124, 131)
• Truncates PR bodies to 200 chars using (.body // "" | .[:200]) with null handling (line 85)
• Directly reads from file in changelog step using jq -r (line 241)

Confidence Score: 5/5

• Safe to merge - solves a critical production issue with a well-tested approach
• The fix addresses a real production failure (ARG_MAX exceeded) with a proven solution (file-based passing). The jq --slurpfile syntax is correct, null body handling prevents edge cases, and the 200-char truncation is appropriate for Claude's use case. All downstream consumers properly read from the file.
• No files require special attention

Important Files Changed

Filename	Overview
.github/workflows/release-on-main.yml	Replaces environment variable passing with file-based approach to avoid ARG_MAX limit, truncates PR bodies to 200 chars

_{Last reviewed commit: 740ca48}