docs: replace restated upstream facts with pointers in discovery, markdown-format #12687

Refs: #7292

Outcome

28 passages in 8 files state a fact an external source owns. Rewrite each as a pointer to the live upstream section, or remove it, so the next upstream change cannot leave a stale copy here. Plugins: discovery, markdown-format.

Done when

• Every row below is fixed: the restated value is deleted, the local decision stays in this repository's own words, and the pointer goes to the exact upstream section with an as-of date and a recheck trigger, per .claude/rules/skill-bodies-state-current-rules.md. A record beside a restated value does not fix it. A row marked Remove has no local dependency on the fact.
• The PR passes check-stamps.sh --trigger-less (plugins/attribution/skills/audit/scripts/check-stamps.sh) on the files it touches.
• Each touched plugin has a changelog fragment.

Rows

discovery

• plugins/discovery/agents/intent-tracer.md:18-20 (r101). Pointer: https://code.claude.com/docs/en/sub-agents#preload-skills-into-subagents Drop the claim that skills preload may not inject the body and its silent-skip behavior; keep the decision to verify the body is present, pointing at the section and the dated record.
• plugins/discovery/agents/intent-tracer.md:42-46 (r102). Pointer: https://code.claude.com/docs/en/sub-agents#preload-skills-into-subagents Drop the statements about forks, conversation view and argument substitution into preloaded bodies; keep the rule not to rely on an unfilled slot, and point at the section.
• plugins/discovery/agents/intent-tracer.md:86-88 (r103). Pointer: https://code.claude.com/docs/en/sub-agents#preload-skills-into-subagents Drop the description of how an unresolved skills entry is skipped and logged; keep the agent's local instruction on what to do if the skill is missing, and point at the section.
• plugins/discovery/agents/intent-tracer.md:103-106 (r104). Pointer: https://code.claude.com/docs/en/sub-agents#available-tools Drop the description of what an omitted tools field inherits; keep that this definition declares no allowlist and has a disallowedTools denylist, and point at available tools.
• plugins/discovery/agents/intent-tracer.md:113-118 (r105). Pointer: https://code.claude.com/docs/en/sub-agents#restrict-which-tools-a-subagent-can-use Drop the claim that an allowlist removes all MCP tools; keep the decision to omit the allowlist because the evidence categories live behind MCP, and point at the section.
• plugins/discovery/agents/intent-tracer.md:220-224 (r106). Pointer: https://code.claude.com/docs/en/sub-agents#frontmatter-reference (confirm the section) Drop the description of what reaching the turn limit does and how turns are counted; keep the local maxTurns value and the instruction to track turns, and point at the frontmatter reference.
• plugins/discovery/agents/intent-tracer.md:320-324 (r107). Pointer: https://code.claude.com/docs/en/sub-agents#resume-subagents (confirm the section) Drop the claims about partial turn-limit output, its text and the version threshold; keep that the disk marker comes first, and point at resuming subagents.
• plugins/discovery/agents/researcher.md:18-25 (r108). Pointer: https://code.claude.com/docs/en/sub-agents#preload-skills-into-subagents Drop the claim about preload failing silently and the debug-log detail; keep the instruction to confirm the skill body is in context first, and point at the section.
• plugins/discovery/agents/researcher.md:59-64 (r109). Pointer: https://code.claude.com/docs/en/skills#available-string-substitutions Drop the description of how the topic does or does not arrive via substitution; keep the rule that a topic absent from the prompt is missing, and point at the section.
• plugins/discovery/agents/researcher.md:83-91 (r110). Pointer: https://code.claude.com/docs/en/skills#available-string-substitutions Drop the explanation of what a substituted effort in a preloaded body holds; keep that source breadth is the caller's written value, and point at the section.
• plugins/discovery/agents/researcher.md:115-120 (r111). Pointer: https://code.claude.com/docs/en/sub-agents#preload-skills-into-subagents Drop the description of unresolved skills being skipped with a debug-log warning; keep the local check that the skill is loaded, and point at the section.
• plugins/discovery/agents/researcher.md:132-144 (r112). Pointer: https://code.claude.com/docs/en/sub-agents#available-tools Drop the statements on how tool pools are inherited and what an allowlist removes; keep the decision to omit a tools allowlist so doc-MCP servers stay available, and point at the section.
• plugins/discovery/agents/researcher.md:266-270 (r113). Pointer: https://code.claude.com/docs/en/sub-agents#frontmatter-reference Drop the description of what reaching the turn limit does; keep this definition's own limit and the stop-gathering discipline, and point at the frontmatter reference for maxTurns.
• plugins/discovery/agents/researcher.md:364-368 (r114). Pointer: https://code.claude.com/docs/en/sub-agents#resume-subagents Drop the claims about partial turn-limit output, resume, and the version-dependent behavior; keep the decision to emit the payload block early as a second channel, and point at the section.
• plugins/discovery/skills/explore/reference/workflow.md:30 (r115). Pointer: https://code.claude.com/docs/en/memory#how-claudemd-files-load Drop the statements on which instruction files are already in context and when path-scoped files load; keep the decision of what to read for conventions, and point at the section.
• plugins/discovery/skills/research/SKILL.md:234 (r117). Pointer: https://code.claude.com/docs/en/skills#available-string-substitutions Drop the description of how the topic does not reach a preloaded body; keep that a dispatched run takes its topic from the dispatch prompt, and point at the section. Same text also at: plugins/discovery/agents/researcher.md:59; a row of its own covers it only if listed.
• plugins/discovery/skills/research/SKILL.md:245-250 (r118). Pointer: https://code.claude.com/docs/en/skills#inject-dynamic-context Drop the claim that the pre-compute block runs as one shell invocation; keep the decision to use separate body Bash calls, and point at the section. Same text also at: plugins/discovery/skills/trace-intent/SKILL.md:21-26; a row of its own covers it only if listed.
• plugins/discovery/skills/research/SKILL.md:315-318 (r119). Pointer: https://code.claude.com/docs/en/workflows#bundled-workflows Drop the description of what deep-research does and its model-invocation setting; keep that the deep tier does not dispatch it and it is for the person to run, and point at the section.
• plugins/discovery/skills/research/SKILL.md:325-327,336-338 (r120). Pointer: https://code.claude.com/docs/en/workflows#bundled-workflows Drop the stated WebSearch requirement and the variation by settings, plan and host; keep the rule that availability is never assumed, and point at the section.
• plugins/discovery/skills/research/context/discipline.md:283,285 (r121). Pointer: https://www.sitemaps.org/protocol.html#sitemaps-xml-format Drop the per-file URL and size caps from the sitemap row and the llms.txt curated-index characterization; keep which file to try first, and point at the sitemaps protocol and https://llmstxt.org/#proposal.
• plugins/discovery/skills/trace-intent/SKILL.md:21-26 (r122). Pointer: https://code.claude.com/docs/en/skills#inject-dynamic-context Drop the claim that the pre-compute block runs as one shell invocation; keep the decision to use separate body Bash calls, and point at the section instead of the repo convention. Same text also at: plugins/discovery/skills/research/SKILL.md:245-250; a row of its own covers it only if listed.
• plugins/discovery/skills/trace-intent/SKILL.md:55-60 (r123). Pointer: https://code.claude.com/docs/en/sub-agents#preload-skills-into-subagents Drop the claim that an unresolved preload is skipped silently and logged only to the debug log; keep the discipline-liveness token requirement, and point at the section. Same text also at: plugins/discovery/agents/researcher.md:115-120; a row of its own covers it only if listed.
• plugins/discovery/skills/trace-intent/context/dispatch.md:107-110 (r124). Pointer: https://code.claude.com/docs/en/sub-agents#resume-subagents Drop the statement that a subagent at its turn limit returns partial output and can be resumed; keep the local turn-limit handling, and point at the section. Same text also at: plugins/discovery/agents/researcher.md:364-368; a row of its own covers it only if listed.

markdown-format

• plugins/markdown-format/README.md:23-29 (r329). Pointer: https://github.com/DavidAnson/markdownlint-cli2#configuration Drop the discovered-name list and the package.json --config behavior; keep the decision that a file is touched only under a discovered config, and point at the section.
• plugins/markdown-format/README.md:23-26 (r331). Pointer: https://github.com/DavidAnson/markdownlint-cli2#configuration Drop the config names and nested-config scoping description; keep the decision that a discovered config opts a file in, and point at the section.
• plugins/markdown-format/README.md:30-36 (r333). Pointer: https://github.com/DavidAnson/markdownlint-cli2#markdownlint-cli2cjs-or-markdownlint-cli2mjs Drop the description of which formats and keys load code through Node; keep the decision that such a config waits for approval, and point at the section.
• plugins/markdown-format/README.md:37-43,181-185 (r330). Pointer: https://github.com/DavidAnson/markdownlint-cli2#markdownlint-cli2jsonc Drop the list of ignore sources git check-ignore consults and the ignores/gitignore behavior claim; keep the decision to use git check-ignore and point at the markdownlint-cli2 section and git-check-ignore docs.
• plugins/markdown-format/README.md:181-185 (r334). Pointer: https://github.com/DavidAnson/markdownlint-cli2#markdownlint-cli2jsonc Drop the claim about which config keys markdownlint-cli2 applies to literal paths and the version stamp; keep the instruction to exempt tracked paths through its own config, and point at the section.

Notes

Lines are as of 3bd42cabfe (https://github.com/melodic-software/claude-code-plugins/blob/3bd42cabfe6bdef37a05778d396869af0381bb51); locate each passage by its text if main has moved. Each rNNN is that row's number in the order of #7292's "Findings by plugin". Every row was upheld by three blind judges and survived a refutation pass, run report-only for #12072. Pointers are candidates: fetch each before writing it, and confirm the section still documents the fact.

Parent: #7292