Exec Error Doctor#

Triage execution failures quickly, classify root cause, and apply targeted fixes instead of random retries.

Quick workflow#

1. Run triage on raw error text:
   • scripts/exec_error_triage.sh <error_text_or_file>
2. Apply the category fix from references/error-taxonomy.md.
3. Re-run with safer wrappers when relevant:
   • GitHub CLI search schema drift: scripts/gh_search_repos_safe.sh
   • ClawHub publish visibility lag: scripts/clawhub_publish_safe.sh
4. Confirm with one clean re-run and capture outcome.

Standard commands#

Generic triage#

bash

bash scripts/exec_error_triage.sh "Unknown JSON field: nameWithOwner"

Safe GitHub repo search (schema-aware)#

bash

bash scripts/gh_search_repos_safe.sh "safe-exec skill" 15

Safe ClawHub publish (retry-aware inspect)#

bash

bash scripts/clawhub_publish_safe.sh ./my-skill my-skill "My Skill" 1.0.0 "Initial release"

Rules#

• Prefer classification before fixes.
• Treat Skill not found right after publish as potentially transient if publish already returned OK.
• For gh search repos --json, prefer fullName (not nameWithOwner).
• Distinguish transient errors (retry/backoff) from hard errors (auth, permission, invalid args).

Resources#

• references/error-taxonomy.md: category→fix map.
• scripts/exec_error_triage.sh: pattern-based classifier.
• scripts/gh_search_repos_safe.sh: resilient GitHub search wrapper.
• scripts/clawhub_publish_safe.sh: publish + retry verification wrapper.