Skip to content

feat(#8330)!: unify repos.yaml config fields across all scopes - #8331

Merged
ggallen merged 4 commits into
mainfrom
agent/8330-unify-repos-config-fields
Oct 11, 2026
Merged

ggallen merged 4 commits into
mainfrom
agent/8330-unify-repos-config-fields

Conversation

@fullsend-ai-coder

Copy link
Copy Markdown
Contributor

Summary

repos.yaml now uses one per-repo config schema at every scope. defaults.config, the new github.config / gitlab.config sections, and each repository entry's config accept every .fullsend/config.yaml field, including runtime and allowed_remote_resources. The old sibling keys and the matching repos install flags are removed. This is a breaking change.

Related Issue

Sub-issue of #7318; unblocks #8219.

Changes

  • Resolution order: code defaults → config_base preset → defaults.config → the repository's forge section config → the repository entry's config (glob entries included). It uses the existing MergeManaged per-field rules:

    • a more specific scalar wins;
    • nested maps merge key by key;
    • a non-empty allowed_remote_resources list is unioned with the inherited list;
    • an explicit empty list replaces the inherited value.

    Each section stays sparse: omitted values are not written into the managed file. github.config never affects GitLab repositories, and gitlab.config never affects GitHub ones.

  • internal/config: ManagedConfig no longer rejects runtime / allowed_remote_resources. ApplyManagedShorthands is removed. New ManagedConfig.ExplicitRuntime() returns a layer's runtime without falling back to code defaults; ResolvedConfig.Runtime uses it so a preset's runtime is not shadowed.

  • internal/repos:

    • Removed RepoEntry / DefaultsConfig Runtime and AllowedRemoteResources, the empty-allowlist marshal shim, and ResolvedConfig.AllowedRemoteResources.
    • Added PlatformConfig.Config.
    • Each layer is validated on its own, with errors naming the section (e.g. github.config.allowed_remote_resources[0]).
    • mergeManagedConfig, RenderManagedConfig and LayeredConfig now take the forge.
    • Adoption guidance moves every existing setting under the entry's config.
  • Old sibling paths rejected at load: defaults.runtime, defaults.allowed_remote_resources, and the repository-entry equivalents fail with the replacement path, for example defaults.runtime is no longer supported; move it to defaults.config.runtime. There is no auto-migration and no alias.

  • CLI:

    • repos install drops --runtime and --allowed-remote-resources; there is no replacement flag.
    • repos set-default rejects defaults.runtime / defaults.allowed_remote_resources with guidance to edit defaults.config, and writes nothing.
  • Docs: docs/cli/repos.md (including a breaking-change upgrade section), docs/architecture.md, docs/reference/config-reference.md, and the repo-management, configuring-gitlab, choosing-a-runtime, operations, layered-config, runtimes and cli-internals guides.

Behavior change to review

A repository-level allowed_remote_resources list used to replace defaults.allowed_remote_resources. Inside config it follows the existing layered merge rule, so a non-empty list is now unioned with the inherited prefixes. [] still denies every remote resource. The issue asks for the existing merge rules, and this is called out in the upgrade docs and in the BREAKING CHANGE trailer.

Testing

  • make lint passes (stage changes first, then run). pre-commit itself could not fetch hook repositories in the sandbox (TLS), so the applicable hooks were run directly and passed: pre-commit-hooks 6.0.0 file checks, gofmt, go vet ./..., gitleaks.
  • Tests added/updated for new or modified logic:
    • New internal/repos/config_platform_layers_test.go. It covers precedence for scalar, nested-map, list, omitted and explicit empty-list values on both forges and on glob-matched repos; sparse rendering; layering above a preset; source-specific validation errors; and rejection of all four old sibling paths.
    • A CLI test checks that dry run, live install and repos status agree on values from defaults.config / github.config.
    • Further tests cover the removed install flags and set-default rejecting the old keys without writing anything.
  • go test ./internal/config ./internal/repos ./internal/gitlablifecycle ./internal/layers passes.
  • go test ./internal/cli has the same six failures as unmodified main in this sandbox: fetch-server socket tests and two post-script env tests.
  • Patch coverage on every touched production function is ≥ 82%.

Checklist

  • PR title follows Conventional Commits (correct type, ! for breaking changes)
  • Commits are signed off (DCO) — human and human-directed agent sessions only (autonomous agent commit; exempt)
  • I wrote this contribution myself and can explain all changes in it

🤖 Generated with Claude Code


Closes #8330

Post-script verification

  • Branch is not main/master (agent/8330-unify-repos-config-fields)
  • Secret scan passed (gitleaks — 9f954d971c4c91fd74a2b1f2b94104127e010503..HEAD)
  • PR body secret scan passed (gitleaks — no-git)

defaults.config, the new github.config / gitlab.config sections, and
each repository entry's config now accept every per-repo
.fullsend/config.yaml field, including runtime and
allowed_remote_resources. Before this change those two fields could
only be set through sibling manifest keys and were rejected inside
config, so operators could not express all per-repo settings in one
schema or use forge-level defaults.

Managed configuration now resolves defaults.config, then the
repository's forge section config, then the entry config (glob entries
included), using the existing per-repo merge rules (MergeManaged): a
more specific scalar wins, maps merge key by key,
allowed_remote_resources unions non-empty lists, and an explicit empty
list replaces the inherited value. Sections stay sparse; omitted
values are never materialized. A forge section's config never affects
the other forge's repositories.

- internal/config: drop the runtime/allowed_remote_resources rejection
  in ManagedConfig and ApplyManagedShorthands; add ExplicitRuntime.
- internal/repos: remove the sibling Runtime/AllowedRemoteResources
  fields and their empty-list marshal shim, add PlatformConfig.Config,
  validate each layer with source-specific paths, and thread the
  platform into mergeManagedConfig, RenderManagedConfig and
  LayeredConfig. Adoption guidance moves every setting under config.
- Manifest loading rejects defaults.runtime,
  defaults.allowed_remote_resources and the repository-entry
  equivalents with the replacement config path. There is no migration
  or alias.
- repos install drops --runtime and --allowed-remote-resources;
  repos set-default rejects the old keys with guidance to edit
  defaults.config and writes nothing.
- Docs: CLI reference, architecture, getting-started, runtime and
  operations guides describe the new schema and the breaking changes.

Verification: config and repos tests pass; internal/cli has the same
six sandbox-environment failures (fetch server socket, post-script env)
as unmodified main. pre-commit could not fetch hook repos (TLS in the
sandbox), so the applicable hooks were run directly (pre-commit-hooks
6.0.0 file checks, gofmt, go vet ./..., gitleaks) and passed.

BREAKING CHANGE: repos.yaml no longer accepts runtime or
allowed_remote_resources as sibling keys under defaults or repository
entries; move them to defaults.config.<field> or the entry's
config.<field>. A repository-level allowed_remote_resources list is
now unioned with inherited prefixes instead of replacing them (use []
to deny all). repos install no longer accepts --runtime or
--allowed-remote-resources, and repos set-default rejects
defaults.runtime and defaults.allowed_remote_resources.

Closes #8330

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:03 PM UTC · Completed 2:16 PM UTC

Commit: cc43813 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $5.30

@codecov

codecov Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 98.05825% with 2 lines in your changes missing coverage. Please review.

Files with missing lines Patch % Lines
internal/repos/manifest.go 95.12% 1 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!

@fullsend-ai-review fullsend-ai-review Bot added the risk/moderate PR risk: moderate label Oct 11, 2026
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

Risk Assessment: moderate (2/5)

Details

Large diff and breaking repos.yaml schema unification with elevated manifest churn are balanced by strong test coverage, no protected or CI workflow edits, complete acceptance criteria coverage, and bot authorship.

Previous run

Risk Assessment: moderate (2/5)

Details

Large diff and breaking schema unification with elevated repo-manifest churn are balanced by strong test coverage, no protected or CI path changes, full acceptance criteria coverage, and bot authorship.

Previous run (2)

Risk Assessment: moderate (2/5)

Details

Large diff and breaking schema unification with elevated repo-manifest churn are balanced by strong test coverage, no protected or CI path changes, full acceptance criteria coverage, and bot authorship.

Previous run (3)

Risk Assessment: moderate (2/5)

Details

Large diff across 36 files with breaking schema changes and high regression churn is counterbalanced by bot authorship, zero protected or security-sensitive paths, high test coverage, and complete alignment with linked issue acceptance criteria.

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

Looks good to me

Earlier findings

  • f_671d839923c289d2 — resolved_by_change: rejectAllowlistBelowReset in internal/repos/managed_config.go now scans from the most specific layer and accepts the chain when the most specific declared allowed_remote_resources list is empty. An empty → non-empty → empty chain therefore resolves to deny-all and renders as []. Two new cases in managed_config_allowlist_test.go cover the accepted chain and the still-rejected empty → non-empty → unset chain.
  • f_a1db3d366d26f3f5 — reclassified (unchanged at info): the exported signature changes are in internal/ packages and the schema break is declared with !. The latest push only loosens a validation rule and adds no new incompatibility.
Previous run

Review

Findings

Low

  • [edge-case] internal/repos/managed_config.go:227 — rejectAllowlistBelowReset returns an error at the first non-empty allowed_remote_resources list that follows an explicit empty list. It never looks at later layers. A chain such as defaults.config: [], github.config: ["https://example.com/"], repository config: [] is therefore rejected, even though the final result is [] and denies every remote resource. This blocks install and status for that repository. The failure is safe (it rejects rather than over-permits) and the case is rare.
    Remediation: Evaluate the whole layer chain before rejecting, so that a later explicit empty list supersedes an earlier reset-then-add sequence. Add a regression test for the empty → non-empty → empty case.

Earlier findings

  • f_773c73c035842045 — resolved_by_change: docs/guides/getting-started/choosing-a-runtime.md line 9 no longer points to a --runtime install flag. It now says to set runtime: in defaults.config, gitlab.config, or the entry config in repos.yaml, and that there is no install flag.
  • f_a1db3d366d26f3f5 — reclassified (low → info): internal/repos cannot be imported from outside the module, and every call site of RenderManagedConfig already passes forgeName. The union behavior follows issue Unify repos.yaml config fields across defaults, platforms, and repositories #8330 and is declared breaking and documented.

Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR
Previous run (2)

Review

Findings

Medium

  • [stale-doc] docs/guides/getting-started/choosing-a-runtime.md:9 — PR feat(#8330)!: unify repos.yaml config fields across all scopes #8331 removes --runtime from fullsend repos install. Line 23 of this page was updated to say GitLab runtime is set in a repos.yaml config section with no install flag, but the intro paragraph on line 9 still says "On GitLab, pass --runtime to fullsend repos install". The page contradicts itself, and following line 9 fails with unknown flag: --runtime.
    Remediation: Rewrite the GitLab sentence on line 9 to match line 23: set runtime: in defaults.config, gitlab.config, or the repository entry's config in repos.yaml before running fullsend repos install.

Low

  • [edge-case] internal/repos/managed_config.go:227 — rejectAllowlistBelowReset errors on the first non-empty allowed_remote_resources list that follows an explicit [], without looking at later layers. The chain defaults.config: [] → github.config: ["https://example.com/"] → repository config: [] is rejected even though the final value is deny-all and could be rendered as []. The failure is conservative and the limitation is documented in docs/cli/repos.md, so this is advisory.
    Remediation: Optionally reject a lost reset only when the most specific declared list is non-empty; add a regression test for the empty → non-empty → empty chain.
  • [backward-incompatible] internal/repos/managed_config.go — A non-empty repo-level allowed_remote_resources used to replace the default list; it now unions with inherited prefixes, which can widen the effective allowlist for migrated manifests. Authorized by Unify repos.yaml config fields across defaults, platforms, and repositories #8330 and documented in docs/cli/repos.md, but the old-key load error does not mention the semantic change, and the docs do not state that a repository can no longer narrow inherited prefixes to a subset.
    Remediation: Optionally mention union semantics in the load error naming the replacement path, and note in docs/cli/repos.md that only [] denies all.

Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR
Previous run (3)

Review

Findings

High

  • [merge-semantics] internal/repos/managed_config.go:72 — Flattening the manifest layers loses an earlier explicit-empty allowlist's inheritance barrier. Example: defaults.config.allowed_remote_resources: [] followed by github.config.allowed_remote_resources: [https://specific.example.com/]. Under the documented semantics this should resolve to only the specific prefix, because the defaults layer already cleared the preset and compiled defaults. Instead MergeManaged/applyManagedLayer flattens [] + [X] to [X] with no reset marker, LayerOnBase then attaches the preset/code-default parents, and AllowedResources() unions their prefixes back in. Resolution and the installed managed file therefore admit prefixes the declared layer chain excluded. The new tests cover an empty final layer but not an empty intermediate layer followed by a non-empty list. See also: [permission-expansion] finding at this location.
    Remediation: Preserve allowlist reset provenance through resolution and managed-file rendering so a more-specific non-empty list cannot bring back prefixes below an earlier explicit-empty layer; if the managed-file schema cannot express this, reject the combination before writing. Add regressions for empty defaults/platform allowlists followed by non-empty child lists, with and without a preset, and verify the installed file's runtime-effective allowlist.

  • [permission-expansion] internal/repos/managed_config.go:71 — Security view of the same flattening defect: with defaults.config.allowed_remote_resources: [] followed by a non-empty github.config/repository/glob list, the effective allowlist is broader than the manifest declares — compiled-in remote prefixes and any config_base prefixes the empty layer should have blocked are re-admitted at runtime. Repository and glob-entry overrides are affected the same way. The established-install safety gate does not reliably catch it (pristine installs skip comparison). See also: [merge-semantics] finding at this location.
    Remediation: Preserve the empty-list inheritance barrier through manifest resolution, managed-file serialization and runtime loading, or reject a non-empty list above an inheritance reset before any write instead of silently broadening access. Add regression tests for defaults/platform [] followed by non-empty platform/repository/glob lists, with both code-default and preset prefixes beneath the reset, covering fresh installs and established-install safety checks.

Medium

  • [documentation] docs/cli/repos.md:521 — Under repos status, the adoption-proposal description still says the proposed repos.yaml entry carries runtime and allowed_remote_resources "as their shorthands", with everything else under config. This PR removes those sibling shorthands and proposedManifestEntry now places every setting under config.
    Remediation: Drop the shorthand wording and state that all existing settings are proposed under the entry's config, matching the updated repo-management guide.

Low

  • [test-weakened] internal/repos/manifest_edit.go:410 — The carved entry now shallow-copies the glob's ManagedConfig, which wraps a shared *perRepoConfig (Writer() returns it directly). Previously AllowedRemoteResources was cloned and TestUpdateInferenceAuth_ConcreteFilterCarvesOutGlob mutated the carved allowlist and asserted the glob was unchanged; the PR removes that isolation assertion rather than translating it to the new Config representation. Impact is latent: current writeback callers replace entry.Config wholesale instead of mutating in place.
    Remediation: Deep-copy ManagedConfig when creating a carved entry and restore a mutation/sibling-isolation regression that uses the config writer.

  • [documentation] docs/guides/getting-started/repo-management.md:131 — Under Inheritance and overrides, the guide still says list fields like allowed_remote_resources "are managed at the defaults level and cannot be cleared per-repo". After this PR it is no longer a defaults sibling key, and inside config it can be cleared per-repo with an explicit empty list (as the same guide's Managed configuration section documents).
    Remediation: Revise or remove the sentence at lines 131-133 so it no longer describes allowed_remote_resources as a top-level list that cannot be cleared per-repo.


Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR

fullsend-ai-review[bot]

This comment was marked as outdated.

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-coder

fullsend-ai-coder Bot commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 Finished Fix · ✅ Success · Started 2:17 PM UTC · Completed 2:29 PM UTC

Commit: cc43813 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $0.60

Reject a non-empty allowed_remote_resources list declared in a more
specific manifest layer than an explicit empty list: the sparse managed
file flattens layers and cannot record the reset, so the preset and code
default prefixes would be re-admitted at runtime. Deep-copy the config
when carving a glob entry into an explicit entry and restore the
isolation regression. Update the repos status and repo-management
documentation for the removed shorthands and the clearable allowlist.

Pre-commit could not fetch hook repos (TLS in sandbox); gofmt, go vet
and the config/repos tests passed. internal/cli has the same six
sandbox-environment failures as before this change.

Addresses #8331

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@fullsend-ai-coder

Copy link
Copy Markdown
Contributor Author

🔧 Fix agent — iteration 1 (bot-triggered)

Addressed all four findings. Chose to reject a non-empty allowlist below an explicit empty one instead of encoding a reset marker, since the sparse managed file schema cannot express it. internal/config and internal/repos tests pass; internal/cli has the same six sandbox-environment failures as before.

Fixed (4):

  1. Flattened manifest layers lose an earlier explicit-empty allowed_remote_resources reset (merge-semantics / permission-expansion) (internal/repos/managed_config.go): The sparse managed file cannot record an inheritance reset, so manifest validation now rejects a non-empty allowlist declared in a more specific layer (platform, repository or glob entry) than an explicit empty one. Added a config accessor for the layer's own declared list, so code defaults are not mixed in, and regression tests for empty-then-non-empty, non-empty-then-empty and all-empty combinations.
  2. Adoption proposal docs still describe runtime and allowed_remote_resources as shorthands (docs/cli/repos.md): Reworded so every proposed setting sits under the entry's config; also documented the new rejection.
  3. Carved glob entry shallow-copies the shared config and the isolation assertion was dropped (internal/repos/manifest_edit.go): Added ManagedConfig.Clone, used it when carving an explicit entry from a glob, and restored a mutation-isolation assertion in the carve-out test.
  4. Guide says list fields like allowed_remote_resources cannot be cleared per-repo (docs/guides/getting-started/repo-management.md): Replaced the sentence with a pointer to the Managed configuration section and documented the empty-list-reset restriction there.

Tests: passed

CI inspections (5):

  1. codecov/patch (passing, success)
  2. test (passing, success)
  3. commit-lint (passing, success)
  4. build (passing, success)
  5. functional-tests (passing, success)

Conflict update: skipped; strategy none; forge MERGEABLE; target main

Next steps:

  • /fs-review — request a re-review of the changes
  • /fs-fix <your instruction> — run another fix pass with specific guidance
  • Push commits directly — review re-runs automatically on push
    Updated by fullsend fix agent

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:31 PM UTC · Completed 2:42 PM UTC

Commit: 5efa7db · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $4.09

fullsend-ai-review[bot]

This comment was marked as outdated.

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-coder

fullsend-ai-coder Bot commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 Finished Fix · ✅ Success · Started 2:44 PM UTC · Completed 2:47 PM UTC

Commit: 5efa7db · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $0.21

- choosing-a-runtime: replace the stale GitLab --runtime install-flag
  sentence with the repos.yaml config instruction.
- repos.md: state that a repository can no longer narrow inherited
  allowed_remote_resources prefixes; only [] denies all.

Not addressed (advisory, optional): relaxing rejectAllowlistBelowReset
for empty->non-empty->empty chains, and extending the old-key load error.
Note: pre-commit could not run (hook repos unreachable from sandbox);
change is docs-only.

Addresses #8331

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@fullsend-ai-coder

Copy link
Copy Markdown
Contributor Author

🔧 Fix agent — iteration 2 (bot-triggered)

Fixed the medium stale-doc finding and the documentation half of the low backward-compat finding. Two optional low-severity code suggestions were declined with reasons. Docs-only change; pre-commit could not run in the sandbox (hook repos unreachable); Go tests were not re-run since no code changed.

Fixed (2):

  1. Stale GitLab --runtime instruction in the intro of choosing-a-runtime.md (docs/guides/getting-started/choosing-a-runtime.md): Rewrote the GitLab sentence to say runtime is set in defaults.config, gitlab.config or the repository entry config in repos.yaml before install, with no install flag.
  2. Docs do not state that a repository can no longer narrow inherited allowed_remote_resources prefixes (docs/cli/repos.md): Added a note that repositories can only add to inherited prefixes and only an explicit empty list denies everything.

Disagreed (2):

  1. rejectAllowlistBelowReset rejects an empty -> non-empty -> empty chain even though the final value is deny-all: Reviewer marked this advisory/optional. The rejection is conservative, safe and already documented in docs/cli/repos.md; relaxing it would add logic to a security-sensitive validator beyond the scope of this review round.
  2. Old-key load error should mention union semantics: Optional; the error is about key relocation and the semantic change is now documented in the migration section of docs/cli/repos.md, so the error text was left unchanged to avoid churn in tests and messages.

Tests: passed

CI inspections (5):

  1. CI / test (passing, success): Passed.
  2. CI / web (passing, success): Passed.
  3. Functional Tests / functional-tests (passing, success): Passed.
  4. E2E Tests / behaviour (passing, success): Passed.
  5. codecov/patch (passing, success): Passed.

Conflict update: skipped; strategy none; forge MERGEABLE; target main

Next steps:

  • /fs-review — request a re-review of the changes
  • /fs-fix <your instruction> — run another fix pass with specific guidance
  • Push commits directly — review re-runs automatically on push
    Updated by fullsend fix agent

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:49 PM UTC · Completed 2:58 PM UTC

Commit: 209bde3 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $2.11

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-coder

fullsend-ai-coder Bot commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 Finished Fix · ✅ Success · Started 3:00 PM UTC · Completed 3:04 PM UTC

Commit: 209bde3 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $0.25

…on PR #8331

rejectAllowlistBelowReset now evaluates the whole layer chain: when the most
specific declared allowed_remote_resources list is empty it supersedes every
earlier layer (deny-all), so an empty -> non-empty -> empty chain is accepted
instead of rejected. Adds regression cases and updates docs/cli/repos.md.

Note: pre-commit could not fetch remote hook repositories (TLS failure); the
applicable local hooks (gofmt, go vet) were run directly and passed.

Addresses #8331

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@fullsend-ai-coder

Copy link
Copy Markdown
Contributor Author

🔧 Fix agent — iteration 3 (bot-triggered)

Addressed the single open low-severity finding by evaluating the whole layer chain before rejecting an allowlist below a reset. Earlier findings were already resolved or reclassified by the reviewer.

Fixed (1):

  1. rejectAllowlistBelowReset rejects an empty, non-empty, empty allowed_remote_resources chain even though the final result is deny-all (internal/repos/managed_config.go): The validator now looks at the most specific declared list first. When that list is empty it supersedes all earlier layers and the chain is accepted. A non-empty most-specific list below an earlier reset is still rejected. Added regression cases (empty, non-empty, empty accepted; empty, non-empty, unset rejected) and clarified docs/cli/repos.md.

Tests: passed

CI inspections (5):

  1. test (passing, success)
  2. functional-tests (passing, success)
  3. commit-lint (passing, success)
  4. codecov/patch (passing, success)
  5. playback (pending, in_progress): E2E playback job still running at inspection time.

Conflict update: skipped; strategy none; forge MERGEABLE; target main

Next steps:

  • /fs-review — request a re-review of the changes
  • /fs-fix <your instruction> — run another fix pass with specific guidance
  • Push commits directly — review re-runs automatically on push
    Updated by fullsend fix agent

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 3:06 PM UTC · Completed 3:14 PM UTC

Commit: ae9a0e6 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.75

@fullsend-ai-review
fullsend-ai-review Bot dismissed stale reviews from themself October 11, 2026 15:14

Superseded by updated review

@fullsend-ai-review fullsend-ai-review Bot added the ready-for-merge All reviewers approved — ready to merge label Oct 11, 2026
@ggallen
ggallen added this pull request to the merge queue Oct 11, 2026
Merged via the queue into main with commit eaa3feb Oct 11, 2026
104 checks passed
@ggallen
ggallen deleted the agent/8330-unify-repos-config-fields branch October 11, 2026 21:25

This branch was previously deployed

1 inactive deployment
dev — ae9a0e6e Deployed Oct 11, 2026 by fullsend-ai-coder[bot] via playback #16302
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ready-for-merge All reviewers approved — ready to merge ready-for-review Triggers review agent dispatch risk/moderate PR risk: moderate

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Unify repos.yaml config fields across defaults, platforms, and repositories

1 participant