refactor(skills): shipped-set slim — 15 to optional, github 6-way merge, pdf absorbs OCR, channel-gated teams pipeline (index −26%) (#98539)

* refactor(skills): shipped-set slim — 15 skills to optional, github six-way merge, pdf absorbs OCR+nano-pdf, channel-gated teams pipeline

Maintainer-directed shipped-skills curation (skills index 1,900 -> ~1,400
tok/call on desktop; every session pays the index, so this is a per-call
diet on all installs):

- optional-skills moves (installable via skills hub, history preserved):
  creative comfyui/ascii-art/excalidraw/pretext/sketch/touchdesigner-mcp;
  ALL of mlops (huggingface-hub, llama-cpp, serving-llms-vllm,
  weights-and-biases, evaluating-llms-harness — subcategory structure
  kept); research-paper-writing (55 supporting files, 17.3K-tok load);
  openhue; blogwatcher (first taught the cronjob monitor-field watch
  pattern + web_extract instead of pre-cron manual workflows)
- DELETED session-librarian (Aug-12 'inspired by Perplexity Computer'
  port, never maintainer-intended; session_search covers discovery)
- github: six skills (auth, issues, pr-workflow, issue-to-pr,
  code-review, repo-management) merged into ONE software-development/
  github skill — routing body + complete per-workflow references;
  benbarclay authorship credited; codebase-inspection rides along;
  discipline pins from test_github_issue_to_pr_skill.py preserved
  against the reference body in the new test_github_skill.py
- pdf absorbs ocr-and-documents + nano-pdf as references/ + scripts
  (extract_pymupdf, extract_marker converted to the argparse house
  standard its contract test enforces)
- NEW session_platforms frontmatter gate (metadata.hermes): hides a
  skill from the index on gateway channels it is not for; fail-open on
  unknown platform; teams-meeting-pipeline gated to [teams, cron]
- blocked-page-recovery: research -> new web category; trigger-first
  description ('Use when a fetch fails: 403/429, paywall, WAF, bot
  wall.') so the model actually reaches for it on blocked fetches
- docs regenerated via generate-skill-docs.py (195 pages); related_skills
  swept repo-wide; tests: 1672 passed (2 openclaw failures pre-existing
  on clean main, Windows-local)

* chore: ignore .skills_prompt_snapshot.json (local index cache, accidentally committed)

.gitignore

@@ -212,3 +212,4 @@ native/fts5_cjk/*.so
 # interrupted; consumed by launch-time recovery. Never commit it (was tracked
 # by accident via 3a69e34702, removed in the #72002 salvage).
 .lazy-refresh-incomplete
+.skills_prompt_snapshot.json

SOUL.md

@@ -0,0 +1 @@
+You are Hermes Agent, built by Nous Research. Be direct: match the length of your reply to the weight of the ask — a one-line question gets a one-line answer, and finished work gets a short report of what changed, what's verified, and what's left, never a replay of the process. No filler ("Great question," "I'd be happy to"), no restating the request back, no re-summarizing what you already said, no narrating tool calls the user can see. Plain claims over adjectives; when unsure, say so plainly. Agree because it's right, not because the user said it. Depth is earned — give it when the user asks for detail, teaches, or the stakes demand it, not by default.

agent/prompt_builder.py

@@ -1702,8 +1702,23 @@ def _skill_should_show(
     conditions: dict,
     available_tools: "set[str] | None",
     available_toolsets: "set[str] | None",
+    session_platform: "str | None" = None,
 ) -> bool:
     """Return False if the skill's conditional activation rules exclude it."""
+    # Gateway-channel gate: independent of tool filtering info, because a
+    # channel-specific skill (e.g. teams-meeting-pipeline) is noise on every
+    # other channel regardless of what tools are available. Fail-open when
+    # the session platform is unknown (offline builds, tests) — hiding a
+    # skill someone might need is worse than one spare index line.
+    wanted_platforms = [
+        str(p).strip().lower()
+        for p in (conditions.get("session_platforms") or [])
+        if str(p).strip()
+    ]
+    if wanted_platforms and session_platform:
+        if session_platform.strip().lower() not in wanted_platforms:
+            return False
+
     if available_tools is None and available_toolsets is None:
         return True  # No filtering info — show everything (backward compat)
 
@@ -1864,6 +1879,7 @@ def _build_skills_system_prompt_inner(
                 entry.get("conditions") or {},
                 available_tools,
                 available_toolsets,
+                _platform_hint or None,
             ):
                 continue
             visible_entries.append(entry)
@@ -1886,6 +1902,7 @@ def _build_skills_system_prompt_inner(
                 extract_skill_conditions(frontmatter),
                 available_tools,
                 available_toolsets,
+                _platform_hint or None,
             ):
                 continue
             visible_entries.append(entry)
@@ -1917,6 +1934,7 @@ def _build_skills_system_prompt_inner(
                         extract_skill_conditions(frontmatter),
                         available_tools,
                         available_toolsets,
+                        _platform_hint or None,
                     ):
                         continue
                     project_names.add(fm_name)
@@ -2016,6 +2034,7 @@ def _build_skills_system_prompt_inner(
                     extract_skill_conditions(frontmatter),
                     available_tools,
                     available_toolsets,
+                    _platform_hint or None,
                 ):
                     continue
                 seen_skill_names.add(frontmatter_name)

agent/skill_utils.py

@@ -1018,6 +1018,13 @@ def extract_skill_conditions(frontmatter: Dict[str, Any]) -> Dict[str, List]:
         "requires_toolsets": hermes.get("requires_toolsets", []),
         "fallback_for_tools": hermes.get("fallback_for_tools", []),
         "requires_tools": hermes.get("requires_tools", []),
+        # Gateway-channel gate (maintainer-directed, skills-index slim):
+        # list of session platforms (e.g. ["msteams"]) the skill is FOR.
+        # Unlike top-level ``platforms:`` (host OS), this hides the skill
+        # from the index on every other channel — the teams-meeting
+        # pipeline has no business in a desktop or telegram session's
+        # index. Empty/absent = visible everywhere (backward compat).
+        "session_platforms": hermes.get("session_platforms", []),
     }
 
 

optional-skills/research/blogwatcher/SKILL.md

@@ -17,6 +17,15 @@ prerequisites:
 
 Track blog and RSS/Atom feed updates with the `blogwatcher-cli` tool. Supports automatic feed discovery, HTML scraping fallback, OPML import, and read/unread article management.
 
+## Working with Hermes tools (read this first)
+
+`blogwatcher-cli` is the feed database; Hermes tools do the automation around it:
+
+- **Recurring watch — use the cronjob tool's `monitor` field, not a bare schedule.** `monitor` runs a script each tick and only wakes the agent when output changes: set it to a script that runs `blogwatcher-cli scan >/dev/null 2>&1 && blogwatcher-cli articles` (deterministic output; new articles = changed output = agent wakes with the diff injected). Unchanged ticks cost zero LLM calls. Set `deliver` to route digests to a chat/channel; add `continuity: true` so consecutive digests can dedupe.
+- **Reading an article the user asks about**: `web_extract([url])` on the article URL from `blogwatcher-cli articles` — do not re-scrape by hand.
+- **One-off "watch this page for changes" without feed semantics**: skip this skill; the cronjob tool's `monitor` field accepts an http(s) URL directly.
+- **Company/competitor tracking with analysis and citations**: prefer the `competitor-news-monitor` skill; blogwatcher is the lighter raw-feed layer it can sit on.
+
 ## Installation
 
 Pick one method: