Skip to content

ci(release): sync version tags to fullsend-ai/agents - #3079

Merged
ggallen merged 3 commits into
fullsend-ai:mainfrom
ggallen:worktree-agents-release-sync
Jul 6, 2026
Merged

ggallen merged 3 commits into
fullsend-ai:mainfrom
ggallen:worktree-agents-release-sync

Conversation

@ggallen

@ggallen ggallen commented Jul 6, 2026 •

Copy link
Copy Markdown
Member

Summary

Closes #3083

  • After GoReleaser completes, the release workflow now pushes the same semver tag to fullsend-ai/agents using an org-owned GitHub App token (RELEASE_APP_ID / RELEASE_APP_PRIVATE_KEY)
  • That tag push triggers agents' own release.yml (see feat: add release workflow for version tag sync agents#25) to create a GitHub Release and move the v0 floating tag
  • Pre-flight and post-flight checklists updated to verify agents repo state

Prerequisites

The fullsend-release GitHub App must be:

  1. Created in the fullsend-ai org with Contents → Read and write permission
  2. Installed on fullsend-ai/agents
  3. App ID stored as RELEASE_APP_ID variable on this repo
  4. Private key stored as RELEASE_APP_PRIVATE_KEY secret on this repo

Test plan

  • Verify RELEASE_APP_ID variable and RELEASE_APP_PRIVATE_KEY secret are set
  • Verify the GitHub App is installed on fullsend-ai/agents
  • Merge companion PR in fullsend-ai/agents (adds release.yml)
  • Cut a release and verify both repos get tagged

🤖 Generated with Claude Code

After GoReleaser completes, the release workflow now pushes the same
semver tag to fullsend-ai/agents using an org-owned GitHub App token.
That tag push triggers agents' own release.yml to create a GitHub
Release and move the v0 floating tag.

Signed-off-by: Claude <noreply@anthropic.com>
Signed-off-by: Greg Allen <gallen@redhat.com>
@qodo-code-review

Copy link
Copy Markdown

PR Summary by Qodo

Sync release semver tags to fullsend-ai/agents after GoReleaser

✨ Enhancement 📝 Documentation ⚙️ Configuration changes 🕐 20-40 Minutes

Grey Divider

AI Description

• Generate an org-owned GitHub App token during releases for cross-repo tag creation.
• Push the same semver tag to fullsend-ai/agents to trigger its release workflow.
• Update release checklists to verify agents repo health and release artifacts.
Diagram

graph TD
  A["Release workflow"] --> B["GoReleaser done"] --> C["Create App token"] --> D["Check tag exists"] --> E["Create tag in agents"] --> F["Agents release.yml"] --> G["Agents Release + move v0"]

  subgraph Legend
    direction LR
    _wf["Workflow step"] ~~~ _api["GitHub API"]
  end
Loading
High-Level Assessment

The following are alternative approaches to this PR:

1. Trigger agents via workflow_dispatch (gh workflow run)
  • ➕ Avoids creating git refs directly in the agents repo
  • ➕ More explicit coupling: trigger a single workflow with parameters
  • ➖ Requires agents workflow to accept dispatch events and handle tag creation internally
  • ➖ Still needs cross-repo auth and careful permission scoping
2. Use a fine-scoped PAT instead of a GitHub App token
  • ➕ Simpler setup than managing App ID + private key
  • ➕ Common operational pattern for small orgs
  • ➖ Harder to scope safely; PAT rotation and ownership are riskier than an org-owned App
  • ➖ Often violates least-privilege expectations for release automation
3. Use a single orchestrator release in agents (agents pulls version from this repo)
  • ➕ One “source of truth” release pipeline
  • ➕ Could centralize release notes and v0 movement logic
  • ➖ More invasive change to release responsibilities and workflow ownership
  • ➖ Introduces new dependency direction and potential failure modes

Recommendation: Keep the PR’s GitHub App + tag-sync approach: it’s minimal, keeps semantic versioning consistent across repos, and leverages native tag-triggered workflows in agents. Ensure the App is tightly scoped (contents RW only, installed only on agents) and consider adding a small guardrail in the future (e.g., validate TAG matches semver and/or restrict to non-fork contexts) to reduce accidental ref creation.

Files changed (4) +77 / -1

Documentation (3) +51 / -1
SKILL.mdAllow gh api/gh pr usage and document agents tag mirroring +7/-1

Allow gh api/gh pr usage and document agents tag mirroring

• Extends the skill’s allowed-tools list to include gh api and gh pr. Documents that releases now also tag fullsend-ai/agents using an org-owned GitHub App token, triggering agents’ own release workflow and v0 movement.

skills/cutting-releases/SKILL.md

post-flight.mdAdd post-flight verification steps for agents release +27/-0

Add post-flight verification steps for agents release

• Adds a checklist section to confirm the agents release workflow ran, the tag/release exists, and the agents v0 floating tag moved for non-prereleases. Provides concrete gh commands for validation and troubleshooting guidance.

skills/cutting-releases/post-flight.md

pre-flight.mdAdd pre-flight checks for agents repo health before cutting a release +17/-0

Add pre-flight checks for agents repo health before cutting a release

• Adds a pre-flight section to review recent workflow runs and open PRs in fullsend-ai/agents prior to releasing. Helps ensure agents is in a releasable state since it will be tagged automatically post-release.

skills/cutting-releases/pre-flight.md

Other (1) +26 / -0
release.ymlMint GitHub App token and mirror release tag to agents repo +26/-0

Mint GitHub App token and mirror release tag to agents repo

• Adds a post-GoReleaser step that generates a GitHub App token using RELEASE_APP_ID/RELEASE_APP_PRIVATE_KEY. Uses gh api to skip if the tag already exists on fullsend-ai/agents, otherwise creates refs/tags/<tag> on agents at the current main SHA.

.github/workflows/release.yml

@fullsend-ai-review

fullsend-ai-review Bot commented Jul 6, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:16 PM UTC · Completed 2:27 PM UTC
Commit: f5c63ba · View workflow run →

@github-actions

github-actions Bot commented Jul 6, 2026 •

Copy link
Copy Markdown

Site preview

Preview: https://ea6a2900-site.fullsend-ai.workers.dev

Commit: 6d1966ea765adbaf975e7de5d676ef578cb091e8

@qodo-code-review

qodo-code-review Bot commented Jul 6, 2026 •

Copy link
Copy Markdown

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📜 Skill insights (0)

Context used
✅ Compliance rules (platform): 54 rules

Grey Divider


Remediation recommended

1. Silent gh api failures ✓ Resolved 🐞 Bug ☼ Reliability
Description
The agents tag-existence check treats any gh api error as “tag does not exist” because it
suppresses stderr and branches only on exit code, so auth/rate-limit/transient failures can lead to
an unnecessary tag-create attempt and a less actionable failure. This can fail the release workflow
after GoReleaser already ran, while hiding the original API error details.
Code

.github/workflows/release.yml[R70-78]

+          if gh api "repos/fullsend-ai/agents/git/ref/tags/${TAG}" >/dev/null 2>&1; then
+            echo "Tag ${TAG} already exists on fullsend-ai/agents, skipping"
+            exit 0
+          fi
+
+          AGENTS_SHA=$(gh api repos/fullsend-ai/agents/git/ref/heads/main --jq '.object.sha')
+          gh api repos/fullsend-ai/agents/git/refs \
+            -f ref="refs/tags/${TAG}" \
+            -f sha="${AGENTS_SHA}"
Relevance

⭐⭐⭐ High

Team previously required explicit gh api failure handling (fail-open/warn) instead of silent
fallbacks; see accepted fixes.

PR-#2398
PR-#1565
PR-#2617

ⓘ Recommendations generated based on similar findings in past PRs

Evidence
The new step suppresses output and treats any non-zero gh api result as “missing tag”, unlike
other workflows in this repo that explicitly handle gh api failures and emit warnings/fallback
behavior.

.github/workflows/release.yml[63-78]
.github/workflows/e2e.yml[111-129]
PR-#2398

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
The release workflow’s agents-tag sync logic uses `if gh api ... >/dev/null 2>&1` to detect tag existence. This makes **404 (missing tag)** indistinguishable from **403/5xx/network** failures, and it also hides diagnostic output.

### Issue Context
This runs after GoReleaser and after moving this repo’s `v0` tag, so an avoidable/opaque failure here can leave the release workflow red even when artifacts were already published.

### Fix Focus Areas
- .github/workflows/release.yml[63-79]

### What to change
- Capture and inspect the HTTP status for the tag lookup (e.g., use `gh api -i` and parse the status line), and:
 - If **200**, tag exists → log and exit 0.
 - If **404**, proceed to create the tag.
 - Otherwise (**403/5xx/etc.**), print the error and fail (or retry with backoff) rather than proceeding as if the tag were missing.
- Don’t fully suppress stderr for unexpected failures; keep enough output to diagnose permission/config/rate-limit problems.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools


Grey Divider

Qodo Logo

Comment thread .github/workflows/release.yml Resolved
@codecov

codecov Bot commented Jul 6, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@fullsend-ai-review

fullsend-ai-review Bot commented Jul 6, 2026 •

Copy link
Copy Markdown

Review

Verdict: comment — clean implementation; protected paths require human approval.

This is a well-scoped addition of cross-repo tag synchronization to the release workflow. The only change since the prior approved review (34ff813) is a 5-line comment block in release.yml that documents the intentional coupling to the v0 step and the pre-release tag sync behavior, addressing both prior low-severity findings. The underlying workflow logic, security posture, and documentation remain sound.

Correctness: The shell script logic is correct. The idempotency check (HTTP status code extraction via --include | head -1 | awk '{print $2}') handles all cases properly: 200 (skip), 404 (create), other codes (error), and pre-HTTP failures (empty HTTP_CODE falls to error branch). The || true is correctly scoped — it prevents set -e from killing the script while the explicit if/elif still enforces error handling. The GitHub API endpoints are correct: GET /git/ref/tags/... (singular) for the existence check, POST /git/refs (plural) for creation.

Security: No vulnerabilities found. All interpolated variables verified individually:

  • HTTP_CODE: constrained to a single whitespace-delimited token from an HTTP status line; cannot contain :: sequences or control characters.
  • TAG / GITHUB_REF_NAME: git ref-name rules forbid : in ref names, preventing workflow command injection. Used only in echo statements and gh api URL paths.
  • AGENTS_SHA: hex SHA from gh api --jq, used only in a -f sha= JSON field.
  • GH_TOKEN: properly scoped via repositories: agents, automatically masked by Actions, not echoed.
  • Trigger: push: tags: v* — only users with write access can push tags. Not fork-accessible.
  • Action pin: create-github-app-token@bcd2ba49218906704ab6c1aa796996da409d3eb1 (v3) — SHA-pinned, consistent with existing workflow patterns.

Intent & scope: Authorized by issue #3083. Scope matches the issue's stated goal (sync version tags to agents repo, update release skill docs). No scope creep. The allowed-tools update in SKILL.md (Bash(gh api:*), Bash(gh pr:*)) is necessary for the new post-flight verification commands.

Documentation: Pre-flight (A2), post-flight (B2), and SKILL.md notes sections are consistent with the workflow behavior and follow existing document structure. No stale documentation found elsewhere in the repo.

Findings

1. Protected paths require human approval — medium

File: multiple | Category: protected-path

This PR modifies files under protected paths that require human reviewer approval:

  • .github/workflows/release.yml
  • skills/cutting-releases/SKILL.md
  • skills/cutting-releases/post-flight.md
  • skills/cutting-releases/pre-flight.md

The PR links to issue #3083 and the description clearly explains the rationale for each change. A human reviewer must approve this PR.

Prior finding resolution

Notes

  • Token scoping: create-github-app-token with repositories: agents generates a token scoped to a single repo with Contents write — appropriate least-privilege for tag creation.
  • Lightweight tag: The API creates a lightweight tag (ref) on agents, while the main repo uses annotated tags. This is fine for triggering the agents release workflow.
  • Companion PR dependency: Pre-flight and post-flight docs reference agents' release.yml being added in feat: add release workflow for version tag sync agents#25. Coordinate merge ordering so agents changes land first.
Previous run

Review

Verdict: approve — clean implementation with only low-severity observations.

Clean, well-scoped addition of cross-repo tag synchronization to the release workflow. Both prior medium and low findings have been addressed: issue #3083 is now linked and the PR title uses the correct ci(release) prefix. The workflow logic is sound — idempotent tag creation with proper set -euo pipefail error handling, SHA-pinned action, and a token scoped to a single repository with least-privilege Contents write. The skill documentation updates (pre-flight, post-flight, SKILL.md notes) are well-placed and consistent with existing structure.

Security: No vulnerabilities found. The ::error:: workflow command interpolates only HTTP_CODE (extracted via awk '{print $2}' from the HTTP status line — a single whitespace-delimited token that cannot contain injection sequences). GITHUB_REF_NAME appears only in plain echo statements and gh api URL paths — git ref-name rules forbid : in ref names, preventing workflow command injection. Token scope is appropriately constrained to repositories: agents.

Findings

1. Implicit step ordering dependency — low

File: .github/workflows/release.yml | Line: 54 | Category: error-handling

Carried from prior review (severity anchored). The "Generate agents repo token" and "Push tag to fullsend-ai/agents" steps have no explicit if: condition, relying on the implicit if: success() default. If the "Move v0 floating tag" step fails unexpectedly, both agents steps are skipped — even though GoReleaser already succeeded and the primary release is published. The agents repo would silently miss the version tag.

Remediation: Either add if: "!cancelled()" to both agents steps so they execute regardless of the v0 step outcome, or add a brief comment confirming that the coupling is intentional.

2. Pre-release tag sync asymmetry — low

File: .github/workflows/release.yml | Line: 54 | Category: design-intent

The v0 floating tag step is deliberately skipped for pre-release tags via if: "!contains(github.ref_name, '-')", but the agents tag-sync steps run for ALL tags including pre-releases (e.g., v1.0.0-rc.1). This asymmetry appears intentional — the post-flight docs distinguish pre-release and non-pre-release verification, and agents' own release.yml would handle pre-release semantics. However, the implicit behavior deserves a brief comment.

Remediation: Add a comment noting that pre-release tags are intentionally synced to agents (or add a matching if guard if they should not be).


Labels: PR modifies CI release workflow and release skill documentation

Previous run

Review

Verdict: approve — clean implementation with only low-severity observations.

Clean, well-scoped addition of cross-repo tag synchronization to the release workflow. Both prior medium and low findings have been addressed: issue #3083 is now linked and the PR title uses the correct ci(release) prefix. The workflow logic is sound — idempotent tag creation with proper set -euo pipefail error handling, SHA-pinned action, and a token scoped to a single repository with least-privilege Contents write. The skill documentation updates (pre-flight, post-flight, SKILL.md notes) are well-placed and consistent with existing structure.

Security: No vulnerabilities found. The ::error:: workflow command interpolates only HTTP_CODE (extracted via awk '{print $2}' from the HTTP status line — a single whitespace-delimited token that cannot contain injection sequences). GITHUB_REF_NAME appears only in plain echo statements and gh api URL paths — git ref-name rules forbid : in ref names, preventing workflow command injection. Token scope is appropriately constrained to repositories: agents.

Findings

1. Implicit step ordering dependency — low

File: .github/workflows/release.yml | Line: 54 | Category: error-handling

Carried from prior review (severity anchored). The "Generate agents repo token" and "Push tag to fullsend-ai/agents" steps have no explicit if: condition, relying on the implicit if: success() default. If the "Move v0 floating tag" step fails unexpectedly, both agents steps are skipped — even though GoReleaser already succeeded and the primary release is published. The agents repo would silently miss the version tag.

Remediation: Either add if: "!cancelled()" to both agents steps so they execute regardless of the v0 step outcome, or add a brief comment confirming that the coupling is intentional.

2. Pre-release tag sync asymmetry — low

File: .github/workflows/release.yml | Line: 54 | Category: design-intent

The v0 floating tag step is deliberately skipped for pre-release tags via if: "!contains(github.ref_name, '-')", but the agents tag-sync steps run for ALL tags including pre-releases (e.g., v1.0.0-rc.1). This asymmetry appears intentional — the post-flight docs distinguish pre-release and non-pre-release verification, and agents' own release.yml would handle pre-release semantics. However, the implicit behavior deserves a brief comment.

Remediation: Add a comment noting that pre-release tags are intentionally synced to agents (or add a matching if guard if they should not be).

Notes


Protected paths detected — this PR modifies files under one or more
protected paths. The review agent cannot approve PRs that touch these paths.
A human reviewer must approve this PR.

Protected files in this PR:

  • .github/workflows/release.yml
  • skills/cutting-releases/SKILL.md
  • skills/cutting-releases/post-flight.md
  • skills/cutting-releases/pre-flight.md

Labels: PR modifies CI release workflow and release skill documentation

Previous run (2)

Review

Verdict: comment — medium-severity findings worth noting, none blocking.

Clean implementation of cross-repo tag synchronization. The workflow logic is sound: idempotent tag creation, proper set -euo pipefail error handling, SHA-pinned action, and token scoped to a single repository. The pre-flight and post-flight documentation updates are well-placed and consistent with the existing structure.

Findings

1. No linked issue for non-trivial change — medium

File: N/A | Category: missing-authorization

This PR adds 77 lines across 4 files (workflow logic, skill docs, pre-flight, post-flight) with no linked issue. Non-trivial changes benefit from a linked issue for traceability — especially for CI infrastructure changes that introduce new secrets/vars (RELEASE_APP_ID, RELEASE_APP_PRIVATE_KEY) and cross-repo write permissions.

Remediation: Create an issue describing the need for agents repo tag synchronization and link it to this PR.

2. PR title prefix — low

File: N/A | Category: commit-type-mismatch

The PR title uses feat(release) but this modifies .github/workflows/release.yml (CI/CD pipeline) and release-process documentation. Per COMMITS.md, feat is reserved for changes "an end user would recognize as new capability." Cross-repo tag synchronization is release infrastructure, not a user-facing feature. The guidance says: "When in doubt, prefer refactor or chore over feat or fix."

Remediation: Consider changing the PR title to ci(release): sync version tags to fullsend-ai/agents.

3. Implicit step ordering dependency — low

File: .github/workflows/release.yml | Line: 54–79 | Category: error-handling

The "Generate agents repo token" and "Push tag to fullsend-ai/agents" steps run unconditionally after "Move v0 floating tag." If the v0 step fails (e.g., an unexpected git push failure), GitHub Actions skips all subsequent steps — meaning the agents tag is NOT pushed even though GoReleaser already succeeded and the primary release is complete. Today the v0 step uses exit 0 for the ancestor-check skip, so this is benign, but the coupling is implicit.

Remediation: If agents tagging should succeed independently of the v0 step, add if: "!cancelled()" to both agents steps. If the coupling is intentional, a brief comment would clarify the design intent.

Notes

  • Pre-release tag handling: The v0 step skips pre-releases (if: "!contains(github.ref_name, '-')") but the agents tag step runs for ALL tags. This appears intentional — the post-flight docs distinguish "For non-prerelease tags, verify the v0 floating tag was moved" — and agents' own release.yml would handle pre-release semantics.
  • Lightweight tag: The API creates a lightweight tag (ref) on agents, while the main repo uses annotated tags. This is fine for triggering the agents release workflow.
  • Token scoping: create-github-app-token defaults owner to the current repo owner (fullsend-ai), so repositories: agents correctly resolves. The token is scoped to a single repo with Contents write — appropriate least-privilege for tag creation.
  • Companion PR dependency: The pre-flight and post-flight docs reference agents' release.yml which is being added in a companion PR. The PR body calls this out — coordinate merge ordering so agents changes land first.

fullsend-ai-review[bot]

This comment was marked as outdated.

The tag-existence check was suppressing all gh api errors, making auth
and rate-limit failures indistinguishable from a missing tag. Now
checks the HTTP status code explicitly: 200 = skip, 404 = create,
anything else = fail with an actionable error.

Signed-off-by: Claude <noreply@anthropic.com>
Signed-off-by: Greg Allen <gallen@redhat.com>
@ggallen ggallen changed the title feat(release): sync version tags to fullsend-ai/agents ci(release): sync version tags to fullsend-ai/agents Jul 6, 2026
@ggallen

ggallen commented Jul 6, 2026

Copy link
Copy Markdown
Member Author

Addressing all three findings:

1. No linked issue — Created #3083 and linked via Closes #3083 in the PR body.

2. PR title prefix — Changed from feat(release) to ci(release). Fair point — this is release infrastructure, not a user-facing feature.

3. Implicit step ordering dependency — Intentional coupling. If the v0 move fails, the release state is suspect and should be investigated before tagging agents. For prereleases, the v0 step is skipped (not failed), so agents steps still run. The inline review finding on this was also resolved with this explanation.

Additionally, the inline finding about silent gh api failures was fixed in 34ff813 — the tag-existence check now inspects the HTTP status code explicitly (200 = skip, 404 = create, anything else = fail with an actionable error).

@fullsend-ai-review

fullsend-ai-review Bot commented Jul 6, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:37 PM UTC · Completed 2:47 PM UTC
Commit: 34ff813 · View workflow run →

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-review fullsend-ai-review Bot added requires-manual-review Review requires human judgment component/ci CI pipelines and checks component/skills and removed requires-manual-review Review requires human judgment labels Jul 6, 2026
Add comments clarifying that agents steps are intentionally coupled
to the v0 step and intentionally run for all tags including
pre-releases.

Signed-off-by: Claude <noreply@anthropic.com>
Signed-off-by: Greg Allen <gallen@redhat.com>
@fullsend-ai-review

fullsend-ai-review Bot commented Jul 6, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:50 PM UTC · Completed 2:59 PM UTC
Commit: 6d1966e · View workflow run →

@fullsend-ai-review fullsend-ai-review Bot added requires-manual-review Review requires human judgment and removed requires-manual-review Review requires human judgment labels Jul 6, 2026

@ralphbean ralphbean left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

@ggallen
ggallen added this pull request to the merge queue Jul 6, 2026
Merged via the queue into fullsend-ai:main with commit 98a7e05 Jul 6, 2026
24 of 25 checks passed
@ggallen
ggallen deleted the worktree-agents-release-sync branch July 6, 2026 17:11
@fullsend-ai-retro

fullsend-ai-retro Bot commented Jul 6, 2026 •

Copy link
Copy Markdown

🤖 Finished Retro · ✅ Success · Started 5:13 PM UTC · Completed 5:20 PM UTC
Commit: 6d1966e · View workflow run →

@fullsend-ai-retro

Copy link
Copy Markdown

Retro: PR #3079 — ci(release): sync version tags to fullsend-ai/agents

This workflow went well overall. Three review iterations were reasonable for a 3-commit human-authored CI change (4 files, 87 additions). The review agent and Qodo provided complementary coverage: Qodo caught a genuine reliability bug (silent gh api failures), while the review agent caught procedural issues (missing issue link, wrong commit type prefix) and delivered a thorough security analysis of variable interpolation and workflow command injection risks. The human reviewer (ralphbean) approved with "LGTM" — no additional findings beyond what agents already surfaced.

Verdict inconsistency (evidence for #2838)

On the second review run (28799582135, commit 34ff813), the review agent produced an approve verdict while simultaneously including a "Protected paths detected — the review agent cannot approve PRs that touch these paths" note. On the very next run (28800468053, commit 6d1966e), the agent correctly produced a comment verdict for the same set of protected paths. The post-script correctly enforced the constraint on both runs (both submitted as COMMENTED, not APPROVED), but the agent's verdict text was wrong on run 2. This is a concrete instance of #2838 — the finding reconciliation step (SKILL.md step 6e-1) may have intermittently downgraded the protected-path finding to info severity, allowing the agent to produce an approve verdict that contradicts the line-926 constraint ("Never approve when any protected-path finding exists, regardless of severity").

Autonomy readiness

The review agent fully covered the human reviewer's feedback on this PR — the human added no findings. The security analysis (variable interpolation, workflow command injection, token scoping) was thorough and correct. However, this class of change (CI release workflows with cross-repo token generation) is appropriately behind the protected-path constraint. The agent's value is in surfacing detailed findings for the human to approve, not in autonomous approval.

No additional proposals needed

Beyond the evidence issue for #2838, no novel improvement opportunities were identified. The workflow executed efficiently with reasonable iteration count, token cost, and time to resolution (~3 hours including human approval wait time).

Proposals filed

This branch was successfully deployed

1 active deployment
site-preview — 6d1966ea Deployed Jul 6, 2026 by github-actions[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

component/ci CI pipelines and checks component/skills requires-manual-review Review requires human judgment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

feat(release): sync version tags to fullsend-ai/agents

2 participants