Skip to content

Universal SEO skill for Claude Code. 26 sub-skills + 19 sub-agents covering technical SEO, E-E-A-T, schema, GEO/AEO, agent readiness (Lighthouse Agentic Browsing, WebMCP, llms.txt), backlinks, local SEO, e-commerce, international SEO, Google APIs, and PDF/Excel reporting. 9 optional extensions for live SEO data.

release: v2.2.6 - security, hosted install, and reliability

Agrici Daniel93c8fc2a103e
diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index 00efd77..2626725 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -5,13 +5,13 @@
"name": "AgriciDaniel"
},
"metadata": {
- "description": "Comprehensive SEO analysis plugin for Claude Code. 25 sub-skills and 18 sub-agents, plus 8 optional MCP extensions (DataForSEO, Firecrawl, Banana, Ahrefs, SE Ranking, Profound, Bing Webmaster, Unlighthouse). v2.2.5 hardens installation, schema validation, rendered-page and accessibility analysis, runtime portability, and current Google Search guidance."
+ "description": "Comprehensive SEO analysis plugin for Claude Code. 25 sub-skills and 18 sub-agents, plus 8 optional MCP extensions (DataForSEO, Firecrawl, Banana, Ahrefs, SE Ranking, Profound, Bing Webmaster, Unlighthouse). v2.2.6 installs from the claude.ai-hosted marketplace, closes a cache-path traversal and a WHOIS referral SSRF, refuses the RFC 6598 range, raises dependency floors, and locks the cost ledgers."
},
"plugins": [
{
"name": "claude-seo",
"source": "./",
- "description": "Comprehensive SEO analysis plugin for Claude Code. 25 sub-skills and 18 sub-agents, plus 8 optional MCP extensions. v2.2.5 hardens installation, schema validation, rendered-page and accessibility analysis, runtime portability, and current Google Search guidance.",
+ "description": "Comprehensive SEO analysis plugin for Claude Code. 25 sub-skills and 18 sub-agents, plus 8 optional MCP extensions. v2.2.6 installs from the claude.ai-hosted marketplace, closes a cache-path traversal and a WHOIS referral SSRF, refuses the RFC 6598 range, raises dependency floors, and locks the cost ledgers.",
"author": {
"name": "AgriciDaniel",
"url": "https://github.com/AgriciDaniel"
diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
index 2fb7f92..773f774 100644
--- a/.claude-plugin/plugin.json
+++ b/.claude-plugin/plugin.json
@@ -1,6 +1,6 @@
{
"name": "claude-seo",
- "version": "2.2.5",
+ "version": "2.2.6",
"description": "Comprehensive SEO analysis plugin for Claude Code. 25 sub-skills (21 core + 1 orchestrator + 1 framework + 2 extension mirrors) and 18 sub-agents cover technical SEO, content quality, schema, sitemaps, Core Web Vitals, local SEO, backlinks, AI/GEO, ecommerce, hreflang, SXO, clustering, drift monitoring, and Google APIs. Includes optional MCP extensions, SPA-aware rendering, portability, and hardened SSRF/DNS-rebinding safe fetchers.",
"author": {
"name": "AgriciDaniel",
diff --git a/.github/workflows/windows-smoke.yml b/.github/workflows/windows-smoke.yml
index 939e2b8..952f3ec 100644
--- a/.github/workflows/windows-smoke.yml
+++ b/.github/workflows/windows-smoke.yml
@@ -18,7 +18,7 @@ on:
target_ref:
description: "Released tag or main branch to install"
required: true
- default: "v2.2.5"
+ default: "v2.2.6"
type: string
permissions:
@@ -37,7 +37,7 @@ jobs:
- id: ref
name: Validate released source ref
env:
- TARGET_REF: ${{ inputs.target_ref || 'v2.2.5' }}
+ TARGET_REF: ${{ inputs.target_ref || 'v2.2.6' }}
shell: bash
run: |
set -euo pipefail
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 9be0efe..d637d09 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,110 +7,72 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
## [Unreleased]
+## [2.2.6] - 2026-09-10
+
+### Security
+
+- `commoncrawl_graph.py` no longer writes outside its cache directory: the `--release`
+ value was interpolated raw into the cache filename, so `--release ../../../../tmp/x`
+ escaped the cache directory and `_save_cache` wrote there. Malformed releases are now
+ rejected at the CLI and path containment is asserted.
+- `domain_history.py` no longer follows an unvalidated WHOIS referral. The IANA
+ `refer:` host is resolved and validated through `url_safety` and dialled at the
+ pinned address; an unusable referral degrades to IANA's own answer.
+- `url_safety.is_safe_ip` now refuses the RFC 6598 shared address space
+ (100.64.0.0/10), where Alibaba Cloud serves instance metadata, and judges
+ IPv4-mapped IPv6 literals by their embedded address. This also refuses Tailscale
+ addresses; see SECURITY.md.
+- WeasyPrint floor raised to 70.0 (PYSEC-2026-3940); requests, google-auth, courlan,
+ playwright, and numpy floors raised to current releases. `pip-audit` passes.
+- SECURITY.md describes only reporting channels that exist: private vulnerability
+ reporting is enabled and the unreachable email fallback is gone.
+
### Added
-### Changed
+- The plugin installs from the claude.ai-hosted marketplace. Hosted sync rejects any
+ plugin with a top-level `bin/` directory, so the launcher moved to
+ `scripts/claude-seo` and every skill, agent, and doc calls it as
+ `"${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo" run <script.py>`. Manual installs copy the
+ launcher to `~/.claude/skills/seo/scripts/claude-seo` and rewrite that token to the
+ absolute path. A layout test keeps `bin/` from coming back. Fixes #298 and #199.
+- CI runs the full suite on Windows and macOS, audits `requirements.txt` with
+ `pip-audit`, and no longer swallows a failed dependency install in the v2 audit.
+- `CLAUDE_SEO_CONFIG_DIR` overrides the config and ledger location, so the ledger tests
+ run against an isolated file. `BANANA_HOME` does the same for the Banana ledger.
+- `preload_check.py --fail-under N` turns the score into an opt-in gate.
+- Regression coverage for the backlink report validator, the schema-hook UTF-8 output,
+ CRLF checkouts, PSI null category scores, ledger concurrency, the hosted plugin layout,
+ and every ledger kind accepted by `seo_updates.py`.
-### Fixed
+### Changed
-- The plugin now installs from the claude.ai-hosted marketplace. Hosted sync
- rejects any plugin that ships a top-level `bin/` directory with
- `marketplace_sync_bin_directory_not_allowed`, which is exactly what the
- launcher lived in, so the listing could never sync. `bin/claude-seo` moved to
- `scripts/claude-seo`, where it resolves `runtime.py` as a sibling, and every
- skill, agent, and doc now calls it through the documented plugin-relative
- form `"${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo" run <script.py>` instead of
- relying on undocumented `bin/` PATH injection. Manual installs are unchanged
- in behaviour: `install.sh` and `install.ps1` copy the launcher to
- `~/.claude/skills/seo/scripts/claude-seo` and rewrite that canonical token to
- the absolute path in the Markdown they install. Fixes #298; supersedes #199.
-- `seo-technical` no longer treats dynamic rendering as a valid setup to verify. Google
- documents it as a workaround rather than a recommended solution, so the audit step now
- flags it as technical debt. Adds a rendering-strategy table (SSR / SSG / CSR) and a
- preferred-framework list so the skill recommends a target state rather than only naming
- the anti-pattern. Detection logic in `fetch_page.py` is unchanged: identifying dynamic
- rendering is still useful, it is only the recommendation that was stale.
-- **The Banana cost ledger had the same defect, with no locking at all.**
- `extensions/banana/scripts/cost_tracker.py` read and wrote `~/.banana/costs.json` with
- no file locking on any platform and a non-atomic write. Measured 4 entries lost out of
- 20 concurrent `log` calls, and 3 of those processes crashed outright with
- `JSONDecodeError` after reading a half-written file. It is also harder to notice than
- the DataForSEO case: this ledger carries running aggregates (`total_cost`,
- `total_images`, `daily`) incremented in the same write as the entry, so a lost write
- drops both together and the file still looks internally consistent. Given the same
- treatment: one exclusive lock across the whole read-modify-write in `log`, atomic
- writes, `msvcrt` fallback with a hard failure when no locking primitive exists, and a
- corrupt ledger that fails closed instead of raising a bare traceback. `reset` takes the
- lock but deliberately does not read first, so it still works as the recovery path for a
- corrupt ledger. `BANANA_HOME` overrides the ledger and pricing paths for the new
- `tests/test_banana_cost_tracker.py`.
-- **Spend ledger lost concurrent writes (`dataforseo_costs.py`).** The ledger took its
- file lock twice, once to read and once to write, and released it in between, so two
- concurrent `log` calls both read the same ledger and the second silently overwrote the
- first. Measured 3 entries lost out of 20 concurrent writes, with 1 to 3 lost on every
- repeat run. This is not platform-specific. It matters because `seo-audit` fans out to
- up to 15 parallel specialists, three of which call DataForSEO and each of which logs
- after every call, so the ledger under-reported most during the runs that spend the
- most. One exclusive lock now spans the entire read-modify-write in both `log` and
- `reset`.
-- **No file locking at all on Windows.** The module set `fcntl = None` when the import
- failed and then took every unlocked code path, so Windows ran the ledger with no
- locking whatsoever. Locking now falls back to `msvcrt`, and when neither primitive is
- available the operation fails loudly rather than proceeding unlocked.
-- **A corrupt ledger silently reset the budget to zero.** The non-atomic
- `open(..., "w")` write could leave a truncated file if the process died mid-write, and
- the reader caught `JSONDecodeError` and returned an empty ledger, which the next write
- then persisted, discarding the whole spend history. Ledger writes are now atomic
- (tempfile plus `os.replace`, matching the idiom already used in `google_auth.py`), and
- an unparseable ledger fails closed with the file left in place for inspection.
-- **`CLAUDE_SEO_CONFIG_DIR`** now overrides the config and ledger location, so the new
- `tests/test_dataforseo_costs.py` concurrency and durability tests run against an
- isolated ledger instead of the operator's real spend history.
-- The Common-Crawl-only no-score rule is now enforced instead of merely stated. The
- `seo-backlinks` skill said not to produce a numeric health score when only Common Crawl
- data is available, while `skills/seo/references/free-backlink-sources.md` told the agent
- to "cap the maximum health score at 70/100" in the same case. The reference gave the more
- actionable of two contradictory instructions, which is the likely reason the rule was
- violated in practice. The guidance now agrees across both files, and
- `validate_backlink_report.py` gains a `source_score_consistency` check that fails a
- report carrying a number with no scoreable source (Moz, Bing, or DataForSEO), or a
- `source: not-assessed` finding carrying a score.
-- `preload_check.py` exited 1 on every successful run that scored below a
- hard-coded 75, so `set -e` scripts and CI steps aborted on healthy pages. A
- completed analysis now exits 0; the gate is opt-in through `--fail-under N`.
- Exit 2 for a URL refused by url_safety is unchanged. (#281)
-- CI now runs the full test suite on Windows and macOS, not just the three
- platform-neutral modules. The tests that assert POSIX mode bits skip on
- Windows instead of failing.
+- `seo-technical` treats dynamic rendering as a workaround to flag, not a target state,
+ and recommends SSR, SSG, or CSR with a preferred-framework list.
+- The `seo-backlinks` skill and `free-backlink-sources.md` no longer contradict each
+ other on Common-Crawl-only reports: no numeric score is produced, and
+ `validate_backlink_report.py` fails a report that carries one.
+- The Repository Topology section of CLAUDE.md describes the two single-remote
+ checkouts and the cherry-pick promotion flow actually in use.
+- The three `test_sync_flow.py` tests that call the live GitHub API run only when
+ `CLAUDE_SEO_NETWORK_TESTS=1`; both CI test jobs set it with `GH_TOKEN`.
### Fixed
-- The schema-hook policy tests decoded the hook's UTF-8 output with the locale
- codec, which is cp1252 on Windows and cannot represent the emoji markers.
-- `consistency_check.py` reported every FLOW-locked prompt file as a hash
- mismatch on a stock Windows clone: Git for Windows checks the files out with
- CRLF (`core.autocrlf=true`) while the lock hashes LF content. The check now
- folds CRLF before hashing, so the lock only fails on real content changes.
-- `url_safety.is_safe_ip` accepted the RFC 6598 shared address space
- (100.64.0.0/10), which Python's `ipaddress` does not count as private. Alibaba
- Cloud serves instance metadata at 100.100.100.200, so a crafted URL or DNS
- answer in that range slipped past the SSRF guard. The range is now refused,
- and IPv4-mapped IPv6 literals are judged as their embedded IPv4 address.
- Note: 100.64.0.0/10 is also the Tailscale range, so audits of a staging site
- reached over Tailscale lose access; see SECURITY.md.
-- `commoncrawl_graph.py` no longer writes outside its cache directory. The
- `--release` value was interpolated raw into the cache filename, so
- `--release ../../../../tmp/x` escaped the cache dir and `_save_cache`'s
- `open(..., "w")` followed it. The value also reaches a download URL, so a
- malformed release is now rejected at the CLI rather than silently rewritten
- (which would leave the cache key disagreeing with what was fetched); path
- components are sanitised and containment asserted as defence in depth.
-- `domain_history.py` no longer follows an unvalidated WHOIS referral. It asks
- IANA for the authoritative server and dialled whatever host came back, over
- plaintext port 43: so anyone able to tamper with that response could point
- it at an internal address and use it to probe the operator's private network.
- The referral is now resolved and validated through `url_safety` and the
- connection is made to the pinned address; an unusable referral degrades to
- IANA's own answer rather than blanking the lookup.
+
+- The DataForSEO and Banana cost ledgers lost concurrent writes and could reset spend
+ history after a truncated write. One exclusive lock now spans each read-modify-write,
+ writes are atomic, Windows falls back to `msvcrt`, and a corrupt ledger fails closed.
+- `pagespeed_check.py` raised `TypeError` when a Lighthouse category returned a null
+ score; the category is now skipped and the page is kept.
+- `preload_check.py` exited 1 on every successful run that scored below 75. A completed
+ analysis exits 0. (#281)
+- `consistency_check.py` reported every FLOW-locked prompt as a hash mismatch on CRLF
+ checkouts; it folds CRLF before hashing.
+- `seo_updates.py --kind documentation` was rejected by argparse while the ledger used
+ that kind; the CLI and the schema now share one list.
+- Optional Google, Bing, and rendering dependencies missing at import time no longer
+ abort the whole pytest session (`pytest.importorskip` in the affected modules).
+- The schema-hook tests decode the hook's UTF-8 output explicitly instead of through the
+ locale codec, which is cp1252 on Windows.
## [2.2.5] - 2026-08-25
diff --git a/CITATION.cff b/CITATION.cff
index 92c2d5b..b283292 100644
--- a/CITATION.cff
+++ b/CITATION.cff
@@ -11,8 +11,8 @@ authors:
repository-code: 'https://github.com/AgriciDaniel/claude-seo'
url: 'https://github.com/AgriciDaniel/claude-seo'
license: MIT
-version: 2.2.5
-date-released: '2026-08-25'
+version: 2.2.6
+date-released: '2026-09-10'
keywords:
- seo
- claude-code
diff --git a/CLAUDE.md b/CLAUDE.md
index 8d7ad1d..2a9e6b1 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -21,7 +21,7 @@ claude-seo/
CONTRIBUTORS.md # Community credits (Pro Hub Challenge)
AGENTS.md # Multi-platform agent instructions (Cursor, Antigravity)
.claude-plugin/
- plugin.json # Plugin manifest (v2.2.5)
+ plugin.json # Plugin manifest (v2.2.6)
marketplace.json # Marketplace catalog for distribution
skills/ # 25 sub-skills (auto-discovered)
seo/ # Main orchestrator skill
diff --git a/README.md b/README.md
index d3f0a0c..ca2c5a4 100644
--- a/README.md
+++ b/README.md
@@ -338,6 +338,7 @@ Test coverage grew from 39 (v1.9.9) to 410 across the v2 line; the url_safety su
- **v2.2.3 (July 2026): prompt-hygiene alignment.** Normalized emphasis and punctuation across the prompt surface without changing behavior, routing, or output contracts.
- **v2.2.4 (July 2026): community maintenance.** Added the managed cross-platform runtime and safe sitemap discovery, repaired GSC pagination and totals, replaced removed Bing endpoints, fixed extension and Windows portability gaps, and reconciled every open issue and pull request.
- **v2.2.5 (August 2026): reliability and Google-currency hardening.** Fixed manual-install data packaging and lxml runtime imports, hardened JSON-LD graphs and rendered-page accessibility analysis, enforced managed-runtime command references, and refreshed Google Search and Lighthouse guidance through August 25.
+- **v2.2.6 (September 2026): security and hosted-install patch.** Closed a cache-path traversal and a WHOIS referral SSRF, refused the RFC 6598 range, moved the launcher to `scripts/` so the claude.ai-hosted marketplace accepts the plugin, raised dependency floors past PYSEC-2026-3940, locked the cost ledgers, and added Windows, macOS, and pip-audit CI.
## Limitations
diff --git a/SECURITY.md b/SECURITY.md
index 7fb86b7..92e466e 100644
--- a/SECURITY.md
+++ b/SECURITY.md
@@ -59,6 +59,8 @@ claude-seo is a research and audit toolkit that runs on a user's workstation. It
- **IPv6-only audit targets.** The strict validator queries `family=AF_INET` for the initial resolution. Hosts with AAAA records only will surface as "DNS resolution failed". This is **fail-closed** by design — we'd rather refuse than connect to an unvalidated IPv6 endpoint. Tracked for a future patch (full dual-stack pinning, similar to the Playwright handler which already uses `AF_UNSPEC`).
- **Windows file permissions.** `os.fchmod(fd, 0o600)` is a no-op on Windows for non-ACL filesystems. Users on Windows should rely on per-user directory ACLs instead of POSIX mode bits.
+- **RFC 6598 (100.64.0.0/10) is refused.** Cloud providers serve instance metadata in this range, so `url_safety` treats it like RFC 1918. Tailscale also allocates from it: a staging site reached over Tailscale cannot be audited with the default policy. There is no escape hatch yet; an explicit local-target allowlist is planned.
+
## Security-relevant code paths
If you are auditing, these are the high-leverage files:
diff --git a/extensions/ahrefs/skills/seo-ahrefs/SKILL.md b/extensions/ahrefs/skills/seo-ahrefs/SKILL.md
index 2819593..967dfb4 100644
--- a/extensions/ahrefs/skills/seo-ahrefs/SKILL.md
+++ b/extensions/ahrefs/skills/seo-ahrefs/SKILL.md
@@ -2,7 +2,7 @@
name: seo-ahrefs
description: Ahrefs API analyst (extension). Reads referring domains, backlinks, organic keywords, and content explorer data via the tested @ahrefs/[email protected] server. Pairs with seo-backlinks for multi-source confidence weighting.
metadata:
- version: "2.2.5"
+ version: "2.2.6"
compatibility: "Tested with @ahrefs/[email protected] (installed by extensions/ahrefs/install.sh)."
---
diff --git a/extensions/banana/skills/seo-image-gen/SKILL.md b/extensions/banana/skills/seo-image-gen/SKILL.md
index f09bf4c..7cc66f7 100644
--- a/extensions/banana/skills/seo-image-gen/SKILL.md
+++ b/extensions/banana/skills/seo-image-gen/SKILL.md
@@ -7,7 +7,7 @@ license: MIT
compatibility: "Requires nanobanana MCP server"
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/extensions/bing-webmaster/skills/seo-bing/SKILL.md b/extensions/bing-webmaster/skills/seo-bing/SKILL.md
index 219089d..5da3617 100644
--- a/extensions/bing-webmaster/skills/seo-bing/SKILL.md
+++ b/extensions/bing-webmaster/skills/seo-bing/SKILL.md
@@ -2,7 +2,7 @@
name: seo-bing
description: Bing Webmaster Tools + IndexNow extension. Microsoft Copilot citations are fed by the Bing index; this skill makes Bing visibility, link data, and IndexNow URL submission first-class.
metadata:
- version: "2.2.5"
+ version: "2.2.6"
compatibility: "Requires BING_WEBMASTER_API_KEY and (optionally) INDEXNOW_KEY in ~/.claude/settings.json env. Run extensions/bing-webmaster/install.sh to configure."
---
diff --git a/extensions/dataforseo/skills/seo-dataforseo/SKILL.md b/extensions/dataforseo/skills/seo-dataforseo/SKILL.md
index a6c62c4..5d5f06a 100644
--- a/extensions/dataforseo/skills/seo-dataforseo/SKILL.md
+++ b/extensions/dataforseo/skills/seo-dataforseo/SKILL.md
@@ -13,7 +13,7 @@ license: MIT
compatibility: "Requires DataForSEO MCP server"
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/extensions/firecrawl/skills/seo-firecrawl/SKILL.md b/extensions/firecrawl/skills/seo-firecrawl/SKILL.md
index 14b4576..6c1a3fb 100644
--- a/extensions/firecrawl/skills/seo-firecrawl/SKILL.md
+++ b/extensions/firecrawl/skills/seo-firecrawl/SKILL.md
@@ -11,7 +11,7 @@ license: MIT
compatibility: "Requires Firecrawl MCP server"
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/extensions/profound/skills/seo-profound/SKILL.md b/extensions/profound/skills/seo-profound/SKILL.md
index dcd8144..ae448e8 100644
--- a/extensions/profound/skills/seo-profound/SKILL.md
+++ b/extensions/profound/skills/seo-profound/SKILL.md
@@ -2,7 +2,7 @@
name: seo-profound
description: Profound LLM citation tracker (extension). Time-series brand citation rates across ChatGPT, Perplexity, and other LLMs. Pairs with seo-seranking for triangulated AI visibility coverage.
metadata:
- version: "2.2.5"
+ version: "2.2.6"
compatibility: "Requires a Profound API key (set PROFOUND_API_KEY by running extensions/profound/install.sh)."
---
diff --git a/extensions/seranking/skills/seo-seranking/SKILL.md b/extensions/seranking/skills/seo-seranking/SKILL.md
index 18f1cba..4b6d44f 100644
--- a/extensions/seranking/skills/seo-seranking/SKILL.md
+++ b/extensions/seranking/skills/seo-seranking/SKILL.md
@@ -2,7 +2,7 @@
name: seo-seranking
description: SE Ranking AI visibility analyst (extension). Tracks AI Share-of-Voice across ChatGPT, Gemini, Perplexity, AI Overviews, and AI Mode in a single query.
metadata:
- version: "2.2.5"
+ version: "2.2.6"
compatibility: "Requires an SE Ranking API key (set SERANKING_API_KEY by running extensions/seranking/install.sh)."
---
diff --git a/extensions/unlighthouse/skills/seo-unlighthouse/SKILL.md b/extensions/unlighthouse/skills/seo-unlighthouse/SKILL.md
index 348bc65..52a3695 100644
--- a/extensions/unlighthouse/skills/seo-unlighthouse/SKILL.md
+++ b/extensions/unlighthouse/skills/seo-unlighthouse/SKILL.md
@@ -2,7 +2,7 @@
name: seo-unlighthouse
description: Multi-page Lighthouse audit via the MIT-licensed Unlighthouse CLI. Free-tier alternative to running PageSpeed against every URL on a site, no API quota burn, runs locally.
metadata:
- version: "2.2.5"
+ version: "2.2.6"
compatibility: "Requires Node 18+ and the unlighthouse npm package. Run extensions/unlighthouse/install.sh to pre-warm."
---
diff --git a/install.ps1 b/install.ps1
index 98336a5..847c999 100644
--- a/install.ps1
+++ b/install.ps1
@@ -118,7 +118,7 @@ $RepoUrl = "https://github.com/AgriciDaniel/claude-seo"
# This default MUST be bumped on every release. CI guard
# (tests/test_manifest_consistency.py) enforces this matches plugin.json.
# Override: $env:CLAUDE_SEO_TAG = 'main'; .\install.ps1
-$RepoTag = if ($env:CLAUDE_SEO_TAG) { $env:CLAUDE_SEO_TAG } else { 'v2.2.5' }
+$RepoTag = if ($env:CLAUDE_SEO_TAG) { $env:CLAUDE_SEO_TAG } else { 'v2.2.6' }
# Create directories
New-Item -ItemType Directory -Force -Path $SkillDir | Out-Null
diff --git a/install.sh b/install.sh
index fa80128..bf4965e 100755
--- a/install.sh
+++ b/install.sh
@@ -12,7 +12,7 @@ main() {
# This default MUST be bumped on every release. CI guard
# (tests/test_manifest_consistency.py) enforces this matches plugin.json.
# Override: CLAUDE_SEO_TAG=main bash install.sh
- REPO_TAG="${CLAUDE_SEO_TAG:-v2.2.5}"
+ REPO_TAG="${CLAUDE_SEO_TAG:-v2.2.6}"
echo "════════════════════════════════════════"
echo "║ Claude SEO - Installer ║"
diff --git a/pyproject.toml b/pyproject.toml
index 327d1a3..512de17 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
[project]
name = "claude-seo"
-version = "2.2.5"
+version = "2.2.6"
description = "Comprehensive SEO analysis skill for Claude Code"
requires-python = ">=3.10"
license = "MIT"
diff --git a/requirements.txt b/requirements.txt
index 378627d..5a94c5a 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,6 +1,6 @@
# Claude SEO - Python Dependencies
# Bounded version pinning with security-conscious minimums
-# Last updated: August 25, 2026 (v2.2.5)
+# Last updated: September 10, 2026 (v2.2.6)
beautifulsoup4>=4.12.0,<5.0.0 # No known CVEs
requests>=2.34.2,<3.0.0 # CVE-2024-47081, CVE-2024-35195 fixes
diff --git a/skills/seo-audit/SKILL.md b/skills/seo-audit/SKILL.md
index c2db1f4..76f4d73 100644
--- a/skills/seo-audit/SKILL.md
+++ b/skills/seo-audit/SKILL.md
@@ -6,7 +6,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-backlinks/SKILL.md b/skills/seo-backlinks/SKILL.md
index d0af1c4..100ad2e 100644
--- a/skills/seo-backlinks/SKILL.md
+++ b/skills/seo-backlinks/SKILL.md
@@ -7,7 +7,7 @@ license: MIT
compatibility: "Free: Common Crawl + verify always available. Optional: Moz API, Bing Webmaster (free signup). Premium: DataForSEO extension."
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-cluster/SKILL.md b/skills/seo-cluster/SKILL.md
index 9b2a181..c7ee0a7 100644
--- a/skills/seo-cluster/SKILL.md
+++ b/skills/seo-cluster/SKILL.md
@@ -14,7 +14,7 @@ license: MIT
metadata:
author: AgriciDaniel
original_author: "Lutfiya Miller (Pro Hub Challenge Winner)"
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-competitor-pages/SKILL.md b/skills/seo-competitor-pages/SKILL.md
index 98a6e73..5cfa689 100644
--- a/skills/seo-competitor-pages/SKILL.md
+++ b/skills/seo-competitor-pages/SKILL.md
@@ -11,7 +11,7 @@ argument-hint: "[url or generate] [competitor]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-content/SKILL.md b/skills/seo-content/SKILL.md
index 197bd27..0665c8a 100644
--- a/skills/seo-content/SKILL.md
+++ b/skills/seo-content/SKILL.md
@@ -9,7 +9,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-dataforseo/SKILL.md b/skills/seo-dataforseo/SKILL.md
index b2be216..49c6e86 100644
--- a/skills/seo-dataforseo/SKILL.md
+++ b/skills/seo-dataforseo/SKILL.md
@@ -13,7 +13,7 @@ license: MIT
compatibility: "Requires DataForSEO MCP server"
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-drift/SKILL.md b/skills/seo-drift/SKILL.md
index a16c330..cb75e45 100644
--- a/skills/seo-drift/SKILL.md
+++ b/skills/seo-drift/SKILL.md
@@ -12,7 +12,7 @@ license: MIT
metadata:
author: AgriciDaniel
original_author: "Dan Colta (Pro Hub Challenge)"
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-ecommerce/SKILL.md b/skills/seo-ecommerce/SKILL.md
index f368935..f712079 100644
--- a/skills/seo-ecommerce/SKILL.md
+++ b/skills/seo-ecommerce/SKILL.md
@@ -14,7 +14,7 @@ compatibility: "Enhanced with DataForSEO Merchant API (optional)"
metadata:
author: AgriciDaniel
original_author: "Matej Marjanovic (Pro Hub Challenge)"
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-flow/SKILL.md b/skills/seo-flow/SKILL.md
index c6ff6a8..200679d 100644
--- a/skills/seo-flow/SKILL.md
+++ b/skills/seo-flow/SKILL.md
@@ -11,7 +11,7 @@ argument-hint: "[stage] [url|topic]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-geo/SKILL.md b/skills/seo-geo/SKILL.md
index c6c6b38..e2916c8 100644
--- a/skills/seo-geo/SKILL.md
+++ b/skills/seo-geo/SKILL.md
@@ -13,7 +13,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-google/SKILL.md b/skills/seo-google/SKILL.md
index 1cc50c7..341bcfe 100644
--- a/skills/seo-google/SKILL.md
+++ b/skills/seo-google/SKILL.md
@@ -12,7 +12,7 @@ argument-hint: "[command] [url|property]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-hreflang/SKILL.md b/skills/seo-hreflang/SKILL.md
index 36ce868..af6b742 100644
--- a/skills/seo-hreflang/SKILL.md
+++ b/skills/seo-hreflang/SKILL.md
@@ -10,7 +10,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-image-gen/SKILL.md b/skills/seo-image-gen/SKILL.md
index a8df5f1..1f6b507 100644
--- a/skills/seo-image-gen/SKILL.md
+++ b/skills/seo-image-gen/SKILL.md
@@ -7,7 +7,7 @@ license: MIT
compatibility: "Requires nanobanana MCP server"
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-images/SKILL.md b/skills/seo-images/SKILL.md
index 51fe21a..fe32d6d 100644
--- a/skills/seo-images/SKILL.md
+++ b/skills/seo-images/SKILL.md
@@ -12,7 +12,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-local/SKILL.md b/skills/seo-local/SKILL.md
index bb162d2..5234288 100644
--- a/skills/seo-local/SKILL.md
+++ b/skills/seo-local/SKILL.md
@@ -13,7 +13,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-maps/SKILL.md b/skills/seo-maps/SKILL.md
index e6eb77e..669df1b 100644
--- a/skills/seo-maps/SKILL.md
+++ b/skills/seo-maps/SKILL.md
@@ -14,7 +14,7 @@ license: MIT
compatibility: "DataForSEO MCP for Tier 1+, Google Maps API for Tier 2"
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-page/SKILL.md b/skills/seo-page/SKILL.md
index 0d9c829..2fce3aa 100644
--- a/skills/seo-page/SKILL.md
+++ b/skills/seo-page/SKILL.md
@@ -10,7 +10,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-plan/SKILL.md b/skills/seo-plan/SKILL.md
index 16fe209..c8bf66b 100644
--- a/skills/seo-plan/SKILL.md
+++ b/skills/seo-plan/SKILL.md
@@ -11,7 +11,7 @@ argument-hint: "[business-type]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-programmatic/SKILL.md b/skills/seo-programmatic/SKILL.md
index fdb05b9..63a8e60 100644
--- a/skills/seo-programmatic/SKILL.md
+++ b/skills/seo-programmatic/SKILL.md
@@ -11,7 +11,7 @@ argument-hint: "[url or plan]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-schema/SKILL.md b/skills/seo-schema/SKILL.md
index bb508e5..b7bdfa8 100644
--- a/skills/seo-schema/SKILL.md
+++ b/skills/seo-schema/SKILL.md
@@ -9,7 +9,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-sitemap/SKILL.md b/skills/seo-sitemap/SKILL.md
index 3b486f5..98f254a 100644
--- a/skills/seo-sitemap/SKILL.md
+++ b/skills/seo-sitemap/SKILL.md
@@ -9,7 +9,7 @@ argument-hint: "[url or generate]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-sxo/SKILL.md b/skills/seo-sxo/SKILL.md
index fbb0bf8..edddeb9 100644
--- a/skills/seo-sxo/SKILL.md
+++ b/skills/seo-sxo/SKILL.md
@@ -13,7 +13,7 @@ license: MIT
metadata:
author: AgriciDaniel
original_author: "Florian Schmitz (Pro Hub Challenge)"
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo-technical/SKILL.md b/skills/seo-technical/SKILL.md
index 534bfe0..1a8e175 100644
--- a/skills/seo-technical/SKILL.md
+++ b/skills/seo-technical/SKILL.md
@@ -10,7 +10,7 @@ argument-hint: "[url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/skills/seo/SKILL.md b/skills/seo/SKILL.md
index 2cd6be0..3929b3e 100644
--- a/skills/seo/SKILL.md
+++ b/skills/seo/SKILL.md
@@ -6,7 +6,7 @@ argument-hint: "[command] [url]"
license: MIT
metadata:
author: AgriciDaniel
- version: "2.2.5"
+ version: "2.2.6"
category: seo
---
diff --git a/tests/test_banana_cost_tracker.py b/tests/test_banana_cost_tracker.py
index 905db47..296302b 100644
--- a/tests/test_banana_cost_tracker.py
+++ b/tests/test_banana_cost_tracker.py
@@ -1,5 +1,5 @@
"""
-Tests for extensions/banana/scripts/cost_tracker.py — cost-ledger integrity.
+Tests for extensions/banana/scripts/cost_tracker.py: cost-ledger integrity.
Same defect class as tests/test_dataforseo_costs.py, and worse here: this
ledger had no file locking at all, non-atomic writes, and running aggregates
diff --git a/tests/test_dataforseo_costs.py b/tests/test_dataforseo_costs.py
index 481f759..faec5c7 100644
--- a/tests/test_dataforseo_costs.py
+++ b/tests/test_dataforseo_costs.py
@@ -1,5 +1,5 @@
"""
-Tests for scripts/dataforseo_costs.py — spend-ledger integrity.
+Tests for scripts/dataforseo_costs.py: spend-ledger integrity.
Regression cover for the lost-update race: the ledger used to take its file
lock twice, once to read and once to write, and drop it in between, so two
@@ -43,7 +43,7 @@ def _entries(config_dir: Path) -> list[dict]:
# ---------------------------------------------------------------------------
-# concurrency — the reported bug
+# concurrency: the reported bug
# ---------------------------------------------------------------------------
diff --git a/tests/test_pagespeed_null_category_score.py b/tests/test_pagespeed_null_category_score.py
index f5f89f3..53c1333 100644
--- a/tests/test_pagespeed_null_category_score.py
+++ b/tests/test_pagespeed_null_category_score.py
@@ -1,7 +1,7 @@
"""Lighthouse emits `score: null` for categories it could not evaluate
(scoreDisplayMode "error" or "notApplicable"). `cat_data.get("score", 0)`
only defaults on a *missing* key, so an explicit null reached the multiply
-and raised TypeError, losing the whole PSI run — including the categories
+and raised TypeError, losing the whole PSI run: including the categories
that did score.
"""