Security, quality, dependency, CI, GitHub, and Codacy scanner hygiene for update-ipsets. MUST be followed when checking project hygiene, GitHub code scanning, GitHub AI findings, Codacy issues/findings, scanner findings, dependency hygiene, CI posture, branch/ruleset enforcement, secret scanning, or supply-chain security.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Security, quality, dependency, CI, GitHub, and Codacy scanner hygiene for update-ipsets. MUST be followed when checking project hygiene, GitHub code scanning, GitHub AI findings, Codacy issues/findings, scanner findings, dependency hygiene, CI posture, branch/ruleset enforcement, secret scanning, or supply-chain security.
Purpose
Project hygiene is a defensive safety net for an AI-generated, highly automated
repository. Treat scanners as primary controls, not as optional advisory noise.
Mandatory Rule
When checking project hygiene, check the full scanner and repository posture.
Resolve both blocking and non-blocking valid findings.
"Not blocking CI" is not a reason to leave a finding unresolved. Every finding
must end in one of these states:
fixed and validated;
explicitly baselined or allowlisted with narrow scope and evidence;
rejected as false positive or non-goal with file/API evidence in the SOW.
Never hide a finding by broad suppression, alert dismissal, or workflow
exclusion unless the SOW records why it is safe.
Do not rely on the user to provide scanner links. Discover the scanner surfaces
yourself from the repository, GitHub, Codacy, CI checks, and local scanner
configuration.
Required Hygiene Surfaces
Check these surfaces when the user asks for project hygiene, scanner setup, or
scanner finding review:
GitHub Security and quality pages, including standard code-quality findings
and AI findings. These are separate from CodeQL/code-scanning alerts and must
be triaged separately.
Dependabot cooldown alignment with package-manager supply-chain policy. For
pnpm v11+ UI dependencies, Dependabot npm updates must not create PRs for
package versions younger than pnpm's active release-age gate unless the SOW
records a narrow emergency exception.
GitHub secret scanning, push protection, non-provider patterns, and local
redacted generic secret scanning.
GitHub Code Scanning visibility for non-GitHub scanners, especially whether
Codacy results are uploaded as SARIF, which workflow owns the upload, which
SARIF category is used, and whether the upload is advisory or required.
Branch protection or repository rulesets for main, including whether
scanner checks are only advisory or actually enforced.
Codacy Cloud repository status: dashboard grade, issue totals, issue
category/severity/language breakdown, current and ignored issues, security
findings, pull-request analysis, quality gate, coverage status, tools,
patterns, coding standard, branch state, and whether findings are advisory or
enforced.
Codacy configuration:
root .codacy.yml / .codacy.yaml for Codacy Cloud path/language/engine
configuration such as exclude_paths, include_paths, languages, and tool
path scopes;
.codacy/codacy.config.json for Codacy Analysis CLI local tool/pattern
configuration;
Cloud tool/pattern changes require Codacy UI/API/CLI import or equivalent
Cloud action. Do not assume committing .codacy/codacy.config.json changes
Cloud analysis.
Root .codacy.yml can configure or exclude paths for a tool, but it cannot
enable or disable tools. If a tool is enabled by a Codacy coding standard,
repository-level tool toggles cannot disable it; change the coding standard
only with explicit SOW evidence because it can affect other repositories.
OpenSSF Scorecard or equivalent supply-chain posture checks.
Local quality/security gates from project-testing, including Go, UI,
nested-tool, vulnerability, static-analysis, race, and strict-test commands
relevant to the changed surface.
Shell and workflow hygiene for tracked scripts/workflows.
Command Pattern
Prefer Makefile targets when they exist. If a target does not exist yet, use
the direct equivalent and record that gap:
make vulncheck
make staticcheck
make golangci-lint
make actionlint
make shellcheck
make gitleaks
go run github.com/rhysd/actionlint/cmd/actionlint@<pinned-version> .github/workflows/*.yml
shellcheck $(git ls-files '*.sh')
go run github.com/zricethezav/gitleaks/v8@<pinned-version> detect --no-banner --redact=100 --source . --exit-code 2
Use GitHub API or gh for repository-side evidence:
workflows and dynamic scanners;
CodeQL default setup and analyses;
code scanning alerts in open, fixed, and dismissed states;
Dependabot alerts;
secret-scanning alerts;
repository security settings;
branch protection and rulesets;
Actions permissions.
Use Codacy Cloud CLI, Codacy Analysis CLI, or the authenticated Codacy UI for
Codacy evidence. If the CLIs are unavailable or unauthenticated, record that
gap and use the browser/UI evidence that is available. Do not write Codacy API
tokens or credentials to durable artifacts.
Finding Handling
Start with open/blocking findings, but continue through warnings,
informational findings, fixed-alert regressions, dismissed alerts, scanner
configuration gaps, and missing enforcement.
For Codacy, start with the issue overview and top patterns by count. Separate
real findings from configuration mismatch before fixing thousands of issues.
Wrong-stack, deprecated-tool, generated-file, vendored-file, fixture, and
project-convention mismatches should be fixed through narrow tool/pattern/path
configuration, not broad suppression.
If Codacy findings are uploaded to GitHub Code Scanning, keep GitHub
visibility and Codacy cleanup decisions separate. Uploading all Codacy SARIF
is acceptable for visibility, but do not disable rules merely to reduce the
GitHub alert count. Rules, patterns, or paths may be removed only after the
SOW records evidence that the rule is irrelevant, duplicated by a better
scanner, or mismatched to this repository.
For structural quality work, do not rely on Codacy's repository-level
complexity or duplication percentages alone as an actionable file map. Run
local tools/archposture for large-file/function evidence and a source-only
jscpd scan for duplicate blocks, excluding generated assets, dependencies,
SOW artifacts, test data, and _test.go files unless the SOW explicitly
targets test maintainability. Record before/after counts in the SOW.
For Codacy file-health work, query the Codacy Files API in addition to
codacy issues. codacy issues can be clean while the Files page still
reports poor complexity, duplication, or coverage. Treat cyclomatic
complexity as the Codacy metric engine in .codacy.yml; lizard
exclusions do not by themselves prove that Codacy file complexity metrics are
excluded. If top file complexity or duplication is dominated by test or
benchmark files, fix the Codacy metric/duplication configuration first and
keep production source files visible.
The official Codacy Analysis CLI GitHub Action runs local analyzers and can
be too slow for this repository when configured to export all findings. For
GitHub Code Scanning visibility, prefer exporting already-computed Codacy
Cloud issues through the Codacy Cloud CLI to SARIF. This requires a GitHub
Actions secret named CODACY_API_TOKEN; never write the token value to repo
files, logs, SOWs, docs, or skills.
For Go file-permission findings, separate test fixtures from production
writers before changing modes. Test fixtures should normally use restrictive
0600 files and 0700 directories. Daemon-owned generated runtime and
publication outputs should also default to 0600 files and 0700
directories because public HTTP availability is served by the daemon, not by
POSIX world-read bits. Keep broader group access only where a documented
install/operator contract requires it, such as root-owned binary/config paths
readable or executable by the group.
Closure Gate
Before closing a hygiene SOW, record:
scanner commands and GitHub API checks run;
Codacy Cloud/UI/CLI checks run, including issue counts before/after and any
tool/pattern/path configuration changes;
every valid blocker and non-blocker outcome;
same-failure searches;
remaining accepted baselines or suppressions with evidence;
whether scanner findings are advisory or enforced;
whether project skills, specs, docs, or AGENTS.md need updates.
iplists
For weak-hash findings, verify the use first. Non-security hashes used for
HTTP cache validators, change tokens, or stable fingerprints are not password
or signature primitives, but prefer SHA-256 anyway when it is cheap and does
not change an operator-facing contract.
For compressed provider/archive extraction findings, fix the reader contract
rather than suppressing the scanner. Gzip, tar/gzip, zip, and similar inputs
must enforce an expanded-byte ceiling, reject entries above the ceiling before
committing output, remove incomplete temp files, and create temp/provider
outputs with private modes unless a documented install contract requires
group access.
When Codacy runs both eslint-8 and eslint-9, treat the repository's
checked-in ESLint v9 flat config as the owner for React/TypeScript UI code.
If legacy eslint-8 is enforced by a coding standard, prefer a narrow
engine-specific .codacy.yml path exclusion for UI paths covered by ESLint9
instead of editing the organization coding standard or disabling security
scanners.
For modern Node .mjs maintenance scripts, do not rewrite ES modules,
async/await, const, or Node builtin imports into obsolete JavaScript only
to satisfy legacy eslint-8 compatibility rules. Add or verify current
ESLint9/root-config coverage for the scripts, then use a narrow eslint-8
path exclusion when the legacy engine is the mismatch. Still fix true script
risks found during triage, such as destructive path cleanup without workspace
bounds, unsafe regexes, shell injection, or untrusted subprocess execution.
For Dependabot npm/pnpm PRs, preserve release-age supply-chain protection.
If CI fails with ERR_PNPM_MINIMUM_RELEASE_AGE_VIOLATION, do not weaken the
installer policy to make the PR pass. Align Dependabot cooldown with pnpm,
wait for the release-age window, or explicitly document a narrow emergency
exception.
Security, Critical, and High findings stay enabled unless another active
scanner/pattern covers the same security concern with better precision. If a
security rule is noisy, prefer file-specific exclusion or a documented false
positive over disabling the whole concern.
For Python subprocess findings, first remove subprocess use when an
in-process project API can do the same work. When subprocess is required for
fixed tools such as git or gh, resolve the executable path with
shutil.which, pass arguments as a list, keep shell=False, set an explicit
check policy, close stdin with subprocess.DEVNULL, and use only narrow
# nosec suppressions with a rationale tied to the fixed command surface.
For shell IFS scanner findings, avoid global or subshell IFS mutation for
simple CSV splitting/joining. Prefer small quoted helper functions that build
arrays or strings explicitly, then verify with ShellCheck and a dry-run path
that exercises the changed argument handling.
Search for the same failure class before committing a fix.
Prefer fixing findings over suppressing them.
If suppression is necessary, make it narrow, durable, and justified by
evidence. Broad path ignores for generated or scratch artifacts are allowed
only when the ignored path is not a source or release input.
Do not write raw secret values or scanner payloads to SOWs, specs, docs,
skills, code comments, or workflow comments. Record only redacted evidence,
rule IDs, paths, line numbers, severities, and states.
