meta fixups

Author: jpcaparas

scripts/validate-all-skills.sh

@@ -38,6 +38,12 @@ while IFS= read -r skill; do
     python3 "${skill}/scripts/test_skill.py" "${skill}"
 done < <(find skills -mindepth 1 -maxdepth 1 -type d -exec test -f "{}/SKILL.md" ';' -print | LC_ALL=C sort)
 
+echo "Checking for leaked builder-only placement metadata"
+if grep -n -E '^## Recommended Destination$' skills/*/SKILL.md; then
+    echo "ERROR: remove builder-only placement sections from shipped SKILL.md files" >&2
+    exit 1
+fi
+
 # Cross-check: files under skills/ that a fresh CI checkout will NOT see.
 # This catches two classes of footgun that each skill's own validate.py
 # cannot detect, because validate.py only inspects the working tree:

skills/eli12/SKILL.md

@@ -123,9 +123,3 @@ Stay in one pass for narrow questions. Extra delegation slows simple explanation
 3. Explaining a subsystem by listing files is not enough. Trace what actually happens.
 4. If the user asked for one narrow thing, do not balloon the answer into a whole-architecture tour.
 5. If the code is messy, say that directly. Accessibility should not sand off real complexity.
-
-## Recommended Destination
-
-Recommended destination: `skills/eli12`
-Reason: this repository already keeps canonical installable skills under `skills/<skill-name>/`.
-Alternative: none needed for this repository.

skills/linkedin-speak/SKILL.md

@@ -82,9 +82,3 @@ Default to the deterministic local translator first. It is reproducible, fast, a
 3. Deterministic output means the same input stays stable across runs. That is useful for tests and memes, but it also means the phrasing can feel formulaic on repeated use.
 4. Kagi's public LinkedIn Speak implementation is not a documented API. This skill uses a local engine by default and only emits a comparison URL for the web UI.
 5. Hashtag selection is keyword-driven. If the input is too vague, the fallback tags will lean generic on purpose.
-
-## Recommended Destination
-
-Recommended destination: `skills/linkedin-speak`
-Reason: this repo is already a public installable skills source that stores canonical skills under `skills/<skill-name>/`.
-Alternative: none needed for this repository.

skills/skill-creator-advanced/SKILL.md

@@ -54,6 +54,8 @@ Reason: <one sentence based on existing skills or current install context>
 Alternative: <optional fallback path if there is a reasonable second choice>
 ```
 
+Treat that recommendation as author-time context only. Do not copy it into the generated skill's `SKILL.md`, `README.md`, `AGENTS.md`, or `metadata.json`.
+
 Classify the skill into a blueprint type:
 
 | Blueprint | When to use | Template |

skills/skill-creator-advanced/references/placement.md

@@ -144,4 +144,6 @@ Reason: this repo already keeps 3 installable skills under .agents/skills
 Alternative: ~/.codex/skills/<skill-name>
 ```
 
+This is conversational guidance for the author before scaffolding. Do not write this block into the generated skill files or repository wrappers.
+
 Then proceed unless the user overrides the destination.