Skills install cleanly under regex skill scanners that block on prompt-injection phrases.
- ads (2.7.4 → 2.7.5): the untrusted-data rule in
references/audit-guardrails.mddescribes an embedded "drop your earlier instructions" directive instead of quoting the literal injection phrase, which Hermes' Skills Guard flagged as a critical prompt injection and blocked on install. Same rule, same meaning. - lead-magnets (2.1.4 → 2.1.5):
references/content-to-skill-corpus-design.mdand eval 9 describe the injected transcript line instead of quoting it, for the same reason. The eval still tests that transcript text is treated as data. - schema: eval 8's expected output rewords "do not ... tell the user to delete" so it no longer reads as an instruction to hide information from the user. Evals only, no version bump.
Skill ZIP assets
Complete source catalog: 53 skills.
Claude web target: 53 skills with reviewed short descriptions.
Perplexity Computer target: 53 skills.
Choose <name>-claude-web.zip for the documented single-folder web upload format.
Choose <name>-perplexity.zip for Perplexity Computer (SKILL.md at the archive root).
Source ZIPs retain original metadata; they are not substitutes for omitted web ZIPs.
Live client upload, account limits, and invocation behavior have not been verified.
Web omissions (also recorded in claude-web-manifest.json):
- None.